BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a complementary metal-oxide semiconductor fabricated as an integrated circuit.

2. Description of the Prior Art

The complementary metal-oxide semiconductor is widely known by the abbreviated term CMOS. The CMOS is a combination of N- and P-channel enhancement-mode devices mounted on a single silicon chip. The advantages thereof include low quiescent power dissipation and high operating speed.

In the conventional CMOS structure, a P channel device is formed in an N^- layer, that is an N-substrate and an N channel device is formed in a P layer which is contained in a part of the N-substrate and is called a P-well. In a highly integrated circuit such devices are separated from one another by no more than a few microns, and a four region PNPN structure is thereby formed in the single silicon chip. This PNPN structure may function as a parasitic silicon-controlled rectifier, that is, an SCR. However, since the SCR remains turned on when external electric noise is applied thereto and maintains a latched status until the anode voltage is removed therefrom, a malfunction occurs in the CMOS or, in the worst case, the CMOS is damaged by a very large short-circuit current flowing through the parasitic SCR. This fact has already been known as the so-called latchup phenomenon.

In the prior art, firstly, a CMOS device which can eliminate the above mentioned latchup phenomenon has been proposed in U.S. Pat. No. 3,955,210. The structural feature of this prior art CMOS is represented by a guard region. Secondly, a CMOS device which can eliminate the above mentioned latchup phenomenon has also been known in the technical field to which the present invention pertains. The structural feature of this latter CMOS is represented by a first heavily doped P⁺ region formed in the P-well, as well as the N⁺ regions which act as source and drain, and a second heavily doped N⁺ region formed in the N-substrate, as well as the P⁺ regions which act as the other source and drain.

The above mentioned two kinds of prior art CMOSs are useful for eliminating the respective latchup phenomena. However, each of these CMOSs has a defect in that the CMOS becomes relatively large in size due to the presence of the above mentioned guard region or the first and second heavily doped regions. Accordingly, it is difficult to fabricate the very highly integrated circuits in the single silicon chip.

SUMMARY OF THE INVENTION

Therefore, it is an object of the present invention to provide a CMOS which can eliminate the latchup phenomenon without increasing the size of the CMOS.

BRIEF DESCRIPTION OF THE DRAWINGS

The present invention will be more apparent from the ensuing description with reference to the accompanying drawings wherein:

FIG. 1 illustrates a cross sectional view of a typical CMOS;

FIG. 2 shows an equivalent circuit diagram of a first parasitic SCR created in a single silicon chip;

FIG. 3 shows an equivalent circuit diagram of a second parasitic SCR created in the single silicon chip;

FIG. 4 illustrates a logic symbol indicating a typical two-input NAND logic gate;

FIG. 5 shows a circuit diagram of P and N channel MOSFETs which are comprised of the two-input NAND logic gate;

FIG. 6 shows a circuit diagram of P and N channel MOSFETs which comprise another logic gate;

FIG. 7 illustrates an actual plan view of a CMOS device which comprise of the two-input NAND logic gate shown in FIGS. 4 and 5, and;

FIG. 8 illustrates a cross sectional view of mainly the MOSFETs Q_p3, Q_p4, Q_n3 and Q_n4 shown in FIG. 6.

DETAILED DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates in cross sectional view a typical CMOS in an N-substrate 11. A P-well 12 is contained in part of the N-substrate 11. A voltage potential V_DD, that is a drain supply voltage, is applied to the N-substrate 11 via an N⁺ region 13 which serves as a contact to the N-substrate 11. Both a P⁺ region 14 and a P⁺ region 15 serve, respectively, as a source and a drain of a P channel MOSFET (MOS fieldeffect transistor) Q_p. The gate of the MOSFET Q_p and conventional insulation layers are not shown in FIG. 1. A voltage potential V_SS, that is a source supply voltage, is applied to the P-well 12 via a P⁺ region 18 which serves as a contact to the P-well 12. Both an N⁺ region 16 and an N⁺ region 17 serve, respectively as a drain and a source of an N channel MOSFET Q_n. The gate of the MOSFET Q_n and conventional insulation layers are not shown in FIG. 1. Thus, the CMOS is comprised of the MOSFETs Q_p and Q_n connected in series, and these MOSFETs are connected between one voltage supply (V_DD) and another voltage supply (V_SS).

Since many P⁺ regions such as the regions 14 and 15 and also other P⁺ regions (not shown) and many N⁺ regions such as the regions 16 and 17 and also other N⁺ regions (not shown), are formed in the N-substrate 11 and the P-well 12, respectively, various kinds of parasitic bipolar devices are created in the single silicon chip on which the MOSFETs Q_p and Q_n and also other logic devices (not shown) are mounted. For example, parasitic PNP and NPN transistors are created in the silicon ship. The parasitic PNP transistors are schematically indicated by symbols of transistors Q_a and Q_c and the parasitic NPN transistors are schematically indicated by symbols of transistors Q_b and Q_d. The reference symbols R₁ and R₂ represent resistive components (resistors) of the N-substrate 11 itself. The reference symbols R₃ and R₄ represent resistive components (resistors) of the P-well 12 itself.

The parasitic PNP transistor Q_a and the parasitic NPN transistor Q_b of FIG. 1 together form a first parasitic circuit having an equivalent circuit diagram as shown in FIG. 2. Also, the parasitic PNP transistor Q_c and the parasitic NPN transistor Q_d of FIG. 1 together form a second parasitic circuit having an equivalent circuit diagram as shown in FIG. 3. Resistors R₅ through R₁₀ shown in FIG. 3 will be explained hereinafter with reference to FIG. 8. Referring to FIG. 2, the emitter of the parasitic transistor Q_a exists in the P⁺ region 14 shown in FIG. 1, and the emitter of the parasitic transistor Q_b exists in the N⁺ region 17 shown in FIG. 1. The base of the transistor Q_a exists in the N-substrate 11 shown in FIG. 1 and the base of the transistor Q_b exists in the P-well 12 shown in FIG. 1. The collector of the transistor Q_a exists in the P-well 12 and the collector of the transistor Q_b exists in the N-substrate 11. In the first parasitic circuit 21, if an external electric noise is applied to the transistor Q_a and the transistor Q_a is slightly turned on due to the external electric noise, a relatively low collector current of the transistor Q_a flows through the resistors R₃ and R₄. Then the transistor Q_b is biased by the voltage developed at the junction of the resistors R₃ and R₄. Therefore, the transistor Q_b is also slightly turned on. Then a relatively low collector current of the transistor Q_b flows through the resistors R₁ and R₂. Accordingly, the transistor Q_b is biased by the voltage developed at the junction of the resistors R₁ and R₂, and the transistor Q_a is caused to be fully turned on, which causes the transistor Q_b to be fully turned on. Thus, both the transistors Q_a and Q_b are finally led to their respective saturation conditions. The saturation conditions will be maintained until the voltage supply (V_DD and/or V_SS) is removed from the first parasitic circuit 21. Thus, the latchup phenomenon occurs in the circuit 21. It should be understood that the latchup phenomenon occurs due to the fact that the circuit 21 is constructed as a parasitic SCR.

Now the technical idea of the present invention for eliminating the parasitic SCR from the single semiconductor chip will be explained in detail. Referring still to FIG. 2, the transistor Q_b is biased by means of the combination of the resistors R₃ and R₄, and similarly the transistor Q_a is biased by means of the combination of the resistors R₁ and R₂. Accordingly, it should be noticed that if the resistor R₄ is removed from the chip, the transistor Q_b can not be turned on. This is because, if there is no such resistor R₄, there is no increase of the voltage level at the base of the transistor Q_b, and thereby the transistor Q_b cannot be biased. Similarly, if the resistor R₁ is removed from the chip, the transistor Q_a cannot be turned on. This is because, if there is no such resistor R₁, there is no decrease of the voltage level at the base of the transistor Q_a, and thereby the transistor Q_a cannot be biased. Since the transistors Q_a and Q_b cooperate with each other for comprising the parasitic SCR, the parasitic SCR may lose its natural latch-up function if at least one of the resistors R₁ or R₄ is removed from the chip. However, it is preferable to remove both resistors R₁ and R₄ therefrom at the same time.

For the purpose of removing the resistors R₁ and/or R₄ from the single silicon chip, in the present invention, additional doped portions are formed therein. The additional doped portions are used for reducing the resistance values of the resistors R₁ and/or R₄ to values being substantially equal to zero. In FIG. 1, since the resistance value of the resistor R₁ may be defined by the length of the distance between the N⁺ region 13 having a voltage potential of V_DD and the parasitic PNP transistor Q_a formed in and close to the P channel MOSFET Q_p, the resistance value of the resistor R₁ can be reduced to a value being substantially equal to zero by introducing a first additional doped portion which is connected to the voltage supply (V_DD), into a part located in or near the source of the P channel MOSFET. In this case, the first additional doped portion must be doped so as to have the same conductivity as that of the N-substrate 11. Similarly, in FIG. 1, since the resistance value of the resistor R₄ may be defined by the length of the distance between the P⁺ region 18 having the voltage potential of V_SS and the parasitic NPN transistor Q_b formed in and close to the N channel MOSFET Q_n, the resistance value of the resistor R₄ can be reduced to a value substantially equal to zero, by introducing a second additional doped portion which is connected to the voltage supply (V_SS), into a part located in the source of the N channel MOSFET. In this case, the second additional doped portion must be doped so as to have the same conductivity as that of the P-well 12.

The above mentioned first and second additional doped portions will be clarified with reference to FIGS. 4, 5, 6 and 7, by taking a two-input NAND logic gate as an example. FIG. 4 illustrates a logic symbol indicating a typical twoinput NAND logic gate. The two-input NAND logic gate can be fabricated by the CMOS which has a circuit diagram as shown in FIG. 5. In FIG. 5, the CMOS is fabricated by P-channel MOSFETs Q_p1 and Q_p2 and N-channel MOSFETs Q_n1 and Q_n2. Each of the P-channel MOSFETs Q_p1 and Q_p2 has substantially the same semiconductor structure as shown by the P-channel MOSFET Q_p in FIG. 1 and also each of the N-channel MOSFETs Q_n1 and Q_n2 has substantially the same semiconductor structure as shown by the N-channel MOSFET Q_n in FIG. 1. The reference symbols IN₁ and IN₂ represent the two inputs (see FIG. 4) of the NAND logic gate. The reference symbol OUT represents the output (see FIG. 4) thereof. In FIG. 5, the above mentioned first additional doped portion is formed at a position, surrounded by a dotted circle 51, corresponding to the source of the MOSFET Q_p1. The identical first additional doped portion is also formed at a position, surrounded by a circle 52, corresponding to the source of the MOSFET Q_p2. On the other hand, the above mentioned second additional doped portion is formed at a position, surrounded by a dotted circle 53, corresponding to the source of the MOSFET Q_n2. Thus, the additional doped portion is entered into at least one MOSFET at the part which is driven by the voltage supply. Preferably, the additional doped portions are entered into the MOSFETs which are driven, at their respective parts, by the voltage supplies (V_DD, V_SS). Therefore, in FIG. 5, the MOSFET Q_n1 does not employ such additional doped portion therein. This is because, the source of the MOSFET Q_n1 is driven by neither the voltage supply (V_DD) nor the voltage supply (V_SS). The reason why it is not necessary for the MOSFET Q_n1 to employ such an additional doped portion will be explained hereinafter with reference to FIGS. 3 and 6. Accordingly, referring to another example, if the present invention is applied to CMOS logic devices having a circuit diagram such as shown in FIG. 6, the first additional doped portion is formed at the source, surrounded by a dotted circle 61, of a P channel MOSFET Q_p3, and the second additional doped portions are formed at respective sources, surrounded by dotted circles 62 and 63, of N channel MOSFETs Q_n4 and Q_n5. Since P channel MOSFETs Q_p4 and Q_p5 are not driven by the voltage supply (V_DD), no such first additional doped portion is required therein. Similarly, since the N channel MOSFET Q_n3 is not driven by the voltage supply (V_SS), no such additional doped portion is required therein. Both P channel MOSFET Q_p6 and N channel MOSFET Q_n6 comprise a conventional output buffer stage.

The present invention will be clarified with reference to FIG. 7. FIG. 7 illustrates in actual plan view a CMOS device comprising the two-input NAND gate shown in FIGS. 4 and 5, including the area of the N-substrate 11 (refer to the N-substrate 11 in FIG. 1) and the area of the P-well 12 (refer to the P-well in FIG. 1). In the N-substrate 11, P⁺ regions 71, 72 and 73 (refer to the P⁺ regions 14 and 15 in FIG. 1) are formed. The P⁺ region 71 serves as the drain D1 of the P channel MOSFET Q_p1 (refer also to FIG. 5). The P⁺ region 72 serves as the source S1 of the MOSFET Q_p1 and, at the same time as the source S2 of the P channel MOSFET Q_p2 (refer also to FIG. 5). The P⁺ region 73 serves as the drain D2 of the MOSFET Q_p2. The gate of the MOSFET Q_p1 is mounted on each part of the p⁺ regions 71 and 72 and also on the area adjacent to both P⁺ regions 71 and 72, which gate 74 is made of polysilicon, and serves as one input IN₁ (refer also to FIGS. 4 and 5). The gate of the MOSFET Q_p2 is mounted on each part of the P⁺ regions 72 and 73 and also on the area adjacent to both P⁺ regions 72 and 73, which gate 75 is made of polysilicon, is and serves as the other input IN₂ (refer also to FIGS. 4 and 5). Output electrode 76 which serves as the output OUT (refer also to FIGS. 4 and 5) is mounted on the P⁺ region 71 and electrically connected thereto via a contact window 77. The output electrode 76 is also mounted on the P⁺ region 73 and electrically connected thereto via a contact window 78. Voltage supply line 79 having the voltage potential of V_DD (refer also to FIG. 5) is mounted on the P⁺ region 72. The voltage potential V_DD is applied to the P⁺ region 72 via a contact window 80.

In the P-well 12, N⁺ regions 81, 82 and 83 (refer to the N⁺ regions 16 and 17 in FIG. 1) are formed. The N⁺ region 81 serves as the drain D1 of the N channel MOSFET Q_n1 (refer also to FIG. 5). The N⁺ region 82 serves as the source S1 of the MOSFET Q_n1 and, at the same, as the drain D2 of the N channel MOSFET Q_n2 (refer also to FIG. 5). The N⁺ region 83 serves as the source S2 of the MOSFET Q_n2. The gate 74 of the MOSFET Q_n1 is mounted on each part of the N⁺ regions 81 and 82 and also on the area adjacent to both N⁺ regions 81 and 82. The gate 75 of the MOSFET Q_n2 is mounted on each part of the N⁺ regions 82 and 83 and also on the area adjacent to both N⁺ regions 82 and 83. The output electrode 76 is mounted on the N⁺ region 81 and is electrically connected thereto via a contact window 87. Voltage supply line 89 having the voltage potential of V_SS (refer also to FIG. 5) is mounted on the N⁺ region 83. The voltage potential V_SS is applied to the N⁺ region 83 via a contact window 90.

In the semiconductor structure of FIG. 7, the above mentioned first and second additional doped portions according to the present invention, are further employed therein. One of the first additional doped portions (see 51 in FIG. 5) is formed as an N region or N⁺ region 91 close to the P channel MOSFET Q_p1 (and also Q_p2). The region 91 has the same conductivity as that of the N-substrate 11 and further is directly connected to the voltage supply line 79 via a contact window 94. Another of the first additional doped portions (see 52 in FIG. 5) is formed as an N region or N⁺ region 92 close to the P channel MOSFET Q_p2 (and also Q_p1). The region 92 has the same conductivity as that of the N-substrate 11 and further is directly connected to the voltage supply line 79 via a contact window 95. Similarly, the second additional doped portion (see 53 in FIG. 5) is formed as a P region or P⁺ region 93 (and 93'). The region 93 (and also 93') has the same conductivity as that of the P-well 12 and further is directly connected to the voltage supply line 89 via a contact window 96 (and also 96'). Thus, the values of the resistors R1 and R4 shown in FIGS. 1 and 2 are reduced to a value substantially equal to zero due to the presence of the first and second additional doped portions, that is the regions 91, 92, 93 and 93'. Further it should be noted that since the additional regions 91, 92, 93 and 93' are not formed far from the P and N channel MOSFETs but close to the same, the CMOS device does not become larger in size compared to the size of typical CMOS device which includes no means for eliminating the latchup phenomenon therefrom.

Furthermore, it should also be noted that since some parts of the N⁺ region 83 and the P⁺ region 72 are cut off so as to form recessed portions therein, and the additional regions 91, 92 and 93, 93' are introduced therein, the size of the CMOS device does not increase compared to the size of a typical CMOS device which includes no means for eliminating the latchup phenomenon therefrom.

As previously mentioned, the MOSFETs which are not driven by the voltage supply (V_DD, V_SS), such as the MOSFET Q_n1 shown in FIG. 5 and the MOSFETs Q_p4, Q_p5, Q_n3 and Q_n5 shown in FIG. 6, require no such first and second additional doped portions to be employed therein. The reason for this fact will be explained by taking the MOSFETs Q_p3, Q_p4, Q_n3 and Q_n4 of FIG. 6 as an example. In FIG. 8, which illustrates a cross sectional view of a semiconductor structure comprising MOSFETs Q_p3, Q_p4, Q_n3 and Q_n4 shown in FIG. 6, the members which are represented by the same reference numerals and symbols shown in FIG. 1, are identical with each other. Resistors R5 and R6 represent resistive components of the N-substrate 11, which resistive components cooperate with the parasitic NPN transistor Q_d, at its collector (refer also to FIG. 1). The transistor Q_d is connected, at its emitter, to the voltage supply (V_SS) via a resistor R7. This emitter corresponds to an N⁺ region acting as the drain of the MOSFET Q_n4. Resistors R9 and R10 represent resistive components of the P-well 12, which resistive components cooperate with the parasitic PNP transistor Q_c at its collector (refer also to FIG. 1). The transistor Q_c is connected, at its emitter, to the voltage supply (V_DD ) via a resistor R8. This emitter corresponds to a P⁺ region acting as the source of the MOSFET Q_p4. The equivalent circuit of the parasitic transistors Q_c, Q_d and the resistors R5 through R10 has already been disclosed in FIG. 3. As seen from FIGS. 2 and 3, the circuit diagram of FIG. 3 is similar to that of FIG. 2, and accordingly, it will be considered, at a glance, that the circuit diagram of FIG. 3 may also constitute the parasitic SCR. However, this parasitic SCR does not create the aforesaid latchup phenomenon. This is because the transistor Q_c is provided with, at its emitter, the resistor R8. Also, the transistor Q_d is provided with, at its emitter, the resistor R7. Resistors which are similar to these resistors R7 and R8, are not presented in the parasitic SCR of FIG. 2. These resistors R7 and R8 function to prevent the circuit of FIG. 3 from operating as the SCR. In a case where an external electric noise is applied to the transistor Q_c and the transistor Q_c is turned on, the collector current thereof flows through the resistors R9 and R10. Then the transistor Q_d is biased and turned on. However, the transistor Q_d is not fully biased due to the increase of the emitter voltage of the transistor Q_d. The increase of the emitter voltage is obtained by a voltage drop developed across the resistor R7 when the emitter current flows therethrough. Thus, the resistor R7 acts to create a negative feedback in the SCR. Similarly, the resistor R8 also acts to create a negative feedback therein. In short, the voltage drop developed across the resistor R5 is cancelled by the voltage drop developed across the resister R8, and accordingly the transistor Q_c cannot fully be biased. As explained above, since the MOSFETs, which are not driven by the voltage supplies, accompany the parasitic transistors Q_c and Q_d (see FIG. 3), no parasitic active SCR is created. Accordingly, no such first and second additional doped portions are required to be employed in such MOSFETs.

According to the present invention, the following three advantages are produced.

(1) A malfunction and damage due to the presence of the latchup phenomenon can be eliminated from a CMOS device.

(2) Said malfunction and damage can be eliminated from a CMOS device without enlarging the size of the CMOS device. This is because, firstly, the additional regions are not formed outside the P and N channel MOSFETs but inside the same, and secondly, recessed portions are formed in the regions of the MOSFETs themselves so as to introduce the additional regions therein.

(3) The CMOS designer can easily design a very highly integrated circuit of CMOS devices. The reason of this fact is as follows. Generally, a great number of P channel MOSFETs are fabricated in the N-substrate of one silicon chip and also a great number of N channel MOSFETs are fabricated in the P-well of the same silicon chip. In the prior art, when the CMOS designer designed the arrangement of said great number of the P and N channel MOSFETs, he had to also design, at the same time, the arrangement of areas which function to eliminate the parasitic SCR, besides the design of the arrangement of the MOSFETs themselves. In recent years, a unique method for determining the layout of logic gates in an integrated circuit has been proposed. In this method, each of the logic gates is formed as a semiconductor cell comprised of MOSFETs. However, in the prior art, when it was required to employ a means for eliminating the latchup phenomenon from the semiconductor cell, the designer had to modify the layout of the logic gates. Contrary to the above, in the present invention, it is not necessary for him to modify the layout of logic gates. That is, in the present invention, it is not necessary for him to design the arrangement of areas which function to eliminate the parasitic SCR, besides the design of the arrangement of the MOSFETs themselves. This is because, in the present invention, such areas which function to eliminate the parasitic SCR can be included in every MOSFET itself independently, as shown by the regions 91, 92 and 93 (93') in FIG. 7.