Semiconductor device having fuse-type memory element.

BACKGROUND OF THE INVENTION

0001The present invention relates to a semiconductor device having a fuse, and more particularly to a programmable memory device having fuse-type memory elements.

0002The fuse-type memory element includes a fusible conductive layer operating as a fuse and represents either one or two states in accordance with its conductibility which is determined by whether or not the fusible conductive layer is broken down. It is required for the fuse-type memory element that the fusible conductive layer is effectively broken down by a programming current. Moreover, it is preferred that the fusible conductive layer is broken down at a predetermined portion thereof.

SUMMARY OF THE INVENTION

0003Therefore, an object of the present invention is to provide a semiconductor device having at least one fuse-type memory element having an improved structure.

0004Another object of the present invention is to provide a semiconductor device having a fuse in which the fuse is reliably broken.

0005A semiconductor device according to the present invention is featured in that a hollow or trench is formed in a thick insulating layer covering a surface of a semiconductor substrate so as to laterally elongate in one direction and a fusible conductive film as a fuse is formed on the surface of the thick insulating film so as to extend in a perpendicular direction and run across the hollow or trench in the thick insulating film.

0006Since the fusible conductive film crosses the hollow formed in the thick insulating film, its thickness is reduced at portions on the edges of the hollow where the steps are formed on the insulating film. Accordingly, the fusible conductive film is effectively broken down at one or both of those portions in response to a programming current, and the broken point or points or the fusible conductive layer is thus predetermined.

0007According to another feature of the present invention, the fusible conductive film crossing the hollow is covered with a reflowable glass layer such as a borophosphosilicate glass layer. The fusible conductive film generates heat when broken down. The glass layer is thereby melted and then fills the broken portion of the fusible conductive layer. Accordingly, the scattering of the fusible conductive film is prevented, and the electrical isolation in the broken down fusible conductive film is maintained.

BRIEF DESCRIPTION OF THE DRAWINGS

0008The above and other objects, features and advantages of the present invention will be more apparent from the following description taken in conjunction with the accompanying drawings, in which <ul id="ul0001" list-style="none"><li>Fig. 1A to Fig. 1I are cross-sectional views in manufacturing steps showing a part of a semiconductor device according to an embodiment of the present invention;</li><li>Fig. 2 is a circuit diagrams of a programmable memory device employing the fuse shown in Fig. 1 as each memory cell;</li><li>Fig. 3 is a partial, cross-sectional view showing a broken down portion of the fuse shown in Fig. 1;</li><li>Fig. 4 is a cross-sectional view showing a device according to another embodiment of the present invention;</li><li>Fig. 5 is a cross-sectional view showing a device according to a still another embodiment of the present invention; and</li><li>Fig. 6 is a circuit diagram of a programmable memory device employing the fuse shown in Fig. 5 as each memory cell.</li></ul>

DETAILED DESCRIPTION OF THE EMBODIMENTS

0009Referring to Figs. 1A to 1I, an embodiment of the present invention will be described below along with its manufacturing steps.

0010As shown in Fig. 1A, a P-type silicon substrate 4 having resistivity of 5 to 10 Ω-cm is prepared. A silicon oxide film 2 having a thickness of about 600 Å and a silicon nitride film 3 having a thickness of about 1000 Å are formed in that order on the substrate 1 and then selectively removed except for transistor formation portions. The nitride film 3 is employed as an oxidation-­resisting film.

0011Referring to Fig. 1B, the substrate 1 is selectively oxidized so that a thick field insulator layer 4 of silicon oxide is selectively formed to a thickness of 5000 to 8000 Å at an exposed surface portion of the substrate 4, partially embedded in the substrate 4.

0012The silicon nitride film 3 is thereafter removed as shown in Fig. 1C.

0013Then, a photoresist layer 5 is formed to cover the entire top surface of the silicon oxide films 2 and 4 and as shown in Fig. 1D, an opening 6 is formed therein to expose the silicon oxide film 2 on the non-oxidized portion 1A of the substrate 1 where a MOS transistor is to be formed. This is a normal, well known step in a process of manufacturing a MOS transistor. According to the present invention, another opening 7 is also formed in the photoresist layer 5 at the same time, to expose a portion of the thick field insulator layer 4. Since the two openings 6 and 7 can be formed simultaneously, there occurs no increase in number of the necessary process steps.

0014Referring to Fig. 1E, etching is performed with the photoresist layer 5 as a mask, to remove the silicon oxide film 2 entirely and at the same time to remove the exposed portion of the field oxide layer 4 to a partial depth. As a result, the transistor formation portion 1A is exposed and a hollow or trench 8 is simultaneously formed in the field oxide layer 4. In order to obtain a desired depth of the hollow 8, etching is continued even after the oxide film 2 disappears. The depth D and width E of the hollow 8 are dependent on a thickness of a fuse film to be formed thereacross. If the depth D of the hollow 8 is larger than the thickness of the fuse film, the fuse film is apt to be separated at the edge of the hollow 8 when formed on the field insulator film. As the hollow 8 becomes shallower, the smaller the reduction in the thickness of the fuse film at the edge of the hollow 8 becomes. If the width E is too small, the fuse film fills the hollow 8 completely and no thickness reduction occurs. Accordingly, relationship of the thickness of the fuse film (represented hereinunder by "T"), the depth D of the hollow 8 and the width E of the hollow 8 is favorably ½T ≦ D ≦ T and E ≧ 2T, respectively. In this embodiment, the fuse film has a thickness T of 5000 Å and the depth D and width E are designed to be 3000 Å and 4 µm, respectively. The length of the hollow 8 is required to be larger than the width of the fuse film.

0015Referring to Fig. 1F, the transistor formation portion 1A is covered with a gate film 9 of silicon oxide having a thickness of 500 Å and formed by thermal oxidation. A polycrystalline silicon layer is then deposited over the entire surface and doped with phosphorus as an N-type impurity. The doped polycrystalline silicon layer is selectively removed to form, a gate electrode 10 of a MOS (Metal-Oxide-Silicon) transistor on the gate film 9 and a fuse film 11 on the thick field insulator layer 4 around the hollow 8. Thus, the doped polycrystalline layer is employed as a fusible conductive layer for the fuse 11 as well as the gate electrode 10. The fuse 11 is elongated in a widthwise direction of the hollow 8 to run across the hollow 8. Accordingly, the thickness of the fuse 11 is reduced at its portions on the edges of the hollow 8 as shown in Fig. 1F. With the above-­mentioned dimensions of the fuse 11 and the hollow 8, the thickness of the fuse film 11 is reduced to about a half at the edges of the hollow 8. The gate electrode 10 and the fuse 11 are thermally oxidized and therefore covered respectively with thin films 12 and 13 of silicon oxide.

0016Arsenic ions or antimory ions as an N-type impurity are selectively implanted into the transistor formation portion 1A of the substrate 1 with the gate electrode 10 and the field insulator 4 as masks, and the anealing is then carried out. N-type drain region 14 and source region 15 are thereby formed as shown in Fig. 1G. Subsequently, an opening 16 is formed in the thin film 13 to expose the upper surface and side surfaces of the portion of the fuse 11 crossing the hollow 8.

0017Referring to Fig. 1H, a borophosphosilicate glass (BPSG) layer 17 as a reflowable glass layer is deposited over the entire surface. The BPSG layer 17 contains boron and phosphorus in an amount of 9.0 mol % (5.0 wt %) and 5.5 mol % (9.5 wt %), respectively. Accordingly, the BPSG layer 17 is melted at a temperature above 850°C. Phosphosilicate glass (PSG) or borosilicate glass (BSG) may be employed as the glass layer 17, but BPSG is superior in stability and reflowability.

0018Referring to Fig. 1I, contact holes are provided in the BPSG layer 17 and the oxide film 9 and 13 to expose a portion of the drain region 14, a portion of the source region 15 and both end portions of the fuse 11. A heat treatment is then carried at temperature of about 900°C for ten minuts to moderate sharpness of the edges of the contact holes provided in the BPSG layer 17. Aluminum is thereafter deposited over the entire surface, and the patterning is then carried out. As a result, a source electrode 18, a connection 19 between the drain region 15 and one end portion of the fuse 11, and an electrode 20 connected to the other end portion of the fuse 11 are formed. A semiconductor device having the fuse 11 is thus manufactured.

0019Referring to Fig. 2, a programmable memory device is shown, as an example of devices employing the fuse 11 of Fig. 1I. The fuse 11 shown in Fig. 1I is represented by "F" and used as a programmable memory element or cell in Fig. 2. A plurality of fuses F₁₁ to F_MN are arranged in a matrix form to constitute a memory cell array 30. Each fuse F is connected between one of word lines W₁ to W_M and one of bit lines B₁ to B_N. The electrodes 19 and 20 shown in Fig. 1I are alongated as the word line W and bit line B, respectively. The word lines W₁ to W_M are connected respectively to N-channel MOS transistors Q_R₁ to Q_RM constituting a row selector 32, the gates thereof being connected to a row decoder 31. The MOS transistor shown in Fig 1I is represented by "Q_R₁" in Fig. 2, for example. The electrode 18 is thus elongated and connected in common to the respective one ends of the transistors Q_R₁ to Q_RM and further to one end of a resistor R₂, the other end thereof being connected to a read-out voltage supply terminal V_R. The one end of the resistor R₂ is also connected to a ground through an N-channel MOS transistor Q_W₀. The bit lines B₁ to B_N are connected in common to one end of a resistor R₁ through N-channel MOS transistors Q_C₁ to Q_CN, respectively. The transistors Q_C₁ to Q_CN constitutes a column selector 32, the gates thereof being connected to a column decoder 35. The other end of the resistor R₁ is grounded. The one end of the resistor R₁ is connected via a sense amplifier 34 to a data output terminal D_OUT. The bit lines B₁ to B_M are further connected in common to a programming voltage supply terminal V_W through N-channel MOS transistors Q_W₁ to Q_WN, respectively, these transistors constituting a write selector 36.

0020The row decoder 31 turns ON one of the transistors Q_R₁ to Q_RM in response to a row address supplied to terminals RA₀ to RA_i. A column address is supplied to terminals CA₀ to CA_j. In data programming mode, the column decoder 35 turns ON one of the transistor Q_W₁ to Q_WN and brings all the transistors Q_C₁ to Q_CN in an OFF state. When the transistors Q_R₁ and Q_W₁ are turned ON in response to row and column address, the fuse F₁₁ is selected. In a case where data "1" is stored in the fuse F₁₁, the terminal V_W is supplied with a programming pulse having a voltage of about 30 V and a current of about 100 mA. The transistor Q_W₀ is turned ON by a signal WE in the data programming mode. The programming current thereby flows into the fuse F₁₁ through the transistor Q_W₁ and reaches the ground via the transistors Q_R₁ and Q_W₀.

0021As already described, the thickness of each fuse F₁₁ is reduced at the edges of the hollow 8 (Fig. 1I). The current capacity is made small in the reduced thickness portions. Accordingly, the fuse 11 is blown off at the edges of the hollow 8 due to the programming current. The electric arc occurs at the blown-off points of the fuse 11. The temperature caused by the electric arc reaches about 1200°C. As a result, the BPSG glass 17 is melted and fills the blown-off portions of the fuse 11 as shown in Fig. 3.

0022As described above, the fuse 11 formed so as to cross the hollow 8 is effectively broken down by the programming current, and the broken points thereof are predetermined to be located at the edges of the hollow 8. Moreover, the cut-off state of the fuse 11 is secured and scattering of the blown-off fuse 11 is prevented by the BPSG glass layer 17.

0023It would be considered that partial reduction of the thickness of the fuse may be attained by selectively etching the fuse from its surface to a desired depth rather than relying upon the hollow 8. However, the selective etching of the fuse film requires additional manufacturing steps. In contrast, according to the present invention, the hollow 8 is formed without increasing the manufacturing steps as described with reference to Fig. 1D and it requires no additional manufacturing steps to make the fuse film 11 selectively thin.

0024If desired, the fuse 11 may be made of other materials such as aluminum than polycrystalline silicon.

0025Turning back to Fig. 2, when data "0" is stored in the selected fuse F, i.e. memory cell, the terminal V_W is applied with 0V. The fuse F does not blow.

0026In a data read mode, the column decoder 35 turns ON one of the transistors Q_C₁ to Q_CN in response to the column address and turns OFF all the transistors Q_W₁ to Q_WN. The signal WE takes the low level to turn the transistor Q_W₀ OFF. When the row decoder 31 and column decoder 35 turns ON the transistors Q_R₁ and Q_C₁, respectively, the fuse F₁₁ is selected. A read-out voltage at the terminal V_R is supplied to one end of the fuse F₁₁ via the resistor R₂ and the transistor Q_R₁. Since the fuse F₁₁ is in a blown state, no current flows. Accordingly, a voltage drop does not occur across the resistor R₁. The sense amplifier 34 invert-amplifies a low level input and produces a high level output, i.e. an output data "1", at the output terminal D_OUT. If the selected fuse F does not blow, the voltage drop occurs across the resistor R₁ and an output data "0" is produced.

0027Fig. 4 shows another embodiment of the present invention in which the same constituents as those in Fig. 1I are denoted by the same reference numerals and their explanation is omitted. In this embodiment, one end of the polycrystalline silicon fuse 11 is extended to the transistor and connected to the drain region 15 of the transistor to function as a drain electrode 40. The source region 14 is also led out by a polycrystalline silicon electrode 41. Since the aluminum electrodes 18 and 19 are not directly contact to the regions 14 and 15, respectively, the so-called alloy spike does not occur to prevent a short-circuit of PN junctions. The regions 14 and 15 are thereby made shallow. Moreover, by utilizing the step of selectively etching the gate oxide film 9 to form the polycrystalline silicon electrodes 40 and 41, the hollow 8 can be made deeper.

0028Referring to Fig. 5, a still another embodiment of the present invention is shown in the form of a bipolar integrated circuit. Formed on a P-type silicon substrate 50 is an N-type epitaxial layer 52. A buried N-type layer 51 of a high impurity concentration is formed between the substrate 50 and the epitaxial layer 52. The epitaxial layer 52 comprises a plurality of island regions 52-1, 52-2, etc. isolated by a P-type isolation regions 53 of a high impurity concentration. The island region 52-1 having the buried layer 51 thereunder operates as a collector region of a bipolar transistor. A field silicon oxide layer 54 partially embedded in the epitaxial layer 52 is selectively formed. Etching of silicon oxide is performed to remove a thin silicon oxide film on the island region 52-1 and at the same time to form a hollow or trench 59 in the field oxide layer 54. A P-type base region 55 and an N-type emitter region 56 are formed in the collector region 52-1. Numerals 57 and 58 denote a collector contact region and a base contact region, respectively. A fuse 60 made of polycrystalline silicon is formed on the field oxide layer 54 to elongate across the hollow 59. One end of the fuse 60 is extended to contact with the emitter region 56 and operates as an emitter electrode 61. A base electrode 62 and a collector electrode 63 both made of polycrystalline silicon are formed on the base contact region 58 and the collector contact region 57, respectively. The electrodes 61 to 63 are separated to one another by silicon oxide films 72 and covered respectively with platinum silicide layers 64 to 66 to reduce the resistivity. A DPSG layer 67 is deposed over the entire surface and contact holes are opened therein. Aluminum electrodes 68 to 71 are formed to contact, respectively, with the other end portion of the fuse 60, the emitter electrode 61, the base electrode 62 and the collector electrode 63.

0029Fig. 6 shows a programmable memory device employing the fuse 60 shown in Fig. 5. In a data programming mode, a terminal 88 is supplied with a data programming power V_W. A row decoder 80 responds to row address data RA₀ to RA_i and turns ON one of transistors QB_R₁ to QB_RM constituting a row selector 81. The transistor shown in Fig. 5 denotes one of the transistors QB_R₁ to QB_RM. Therefore, the electrode 71 is elongated to connect the terminal 88 and the electrode 70 is elongated to connect the row decoder 80. A memory cell array 82 includes a purality of fuses F₁₁ to F_MN arranged in a matrix form, one of which is denoted by the fuse 60 in Fig. 5. Therefore, the electrode 69 is elongated as one of word lines W₁ to W_M and the electrode 68 is elongated as one of bit lines B₁ to B_N. A column decoder 87 responds to column address data CA₀ to CA_j turns ON one of transistors QB_W₁ to QB_WN constituting a write selector 83. On the other hand, all transistors QB_C₁ to QB_CN constituting a column selector 84 are maintained in an OFF state. When the transistors QB_R₁ and QB_W₁ are turned ON, the fuse F₁₁ is selected and supplied with the programming power. The fuse F₁₁ thereby blows on the edge of the hollow 59 (Fig. 5) and the BPSG layer 67 fills the blown section.

0030In a data read operation, the terminal 88 is supplied with a read-out voltage V_R. One of the transistors QB_C₁ to QB_CN is turned ON in response to the column address data, and all the transistors QB_W₁ to QB_WN are maintained in the OFF state. If the selected fuse F is in the blown state, no current flows the bit line B. Therefore, a constant current source 85 draws a predetermined current only from the base-emitter junction of the turned-ON transistor QB_C. The voltage drop across a base resistor R_C is thereby made large, so that the emitter potential of the turned-ON transistor QB_C is decreased largely. A sense amplifier 86 invert-amplifies its input level and produce an output data D_OUT of "1". If the selected fuse F does not blow, the current flowing through the resistor R_C is made small, so that the decrease in emitter potential of the transistor QB_C is small. Therefore, the sense amplifier 86 produces an output data D_OUT of "0".

0031The present invention is not limited to the above embodiments, but may be modified and changed without departing form the scope and spirit of the invention. In particular, the number of the fuses used in a semiconductor device may be at least one.