Process of Fabricating Semiconductor Device Having Doped Polysilicon Layer Without Segregation of Dopant Impurity

This invention relates to a semiconductor device and, more particularly, to a process of fabricating a semiconductor device having a lamination of doped polysilicon and refractory metal silicide used as a wiring and an electrode.

Doped polysilicon finds a wide variety of application such as, for example, an gate electrode of a field effect transistor, a bit line and a stacked capacitor in a dynamic random access memory device. The doped polysilicon per se is electrically conductive, and forms the gate electrode, the bit line and the stacked capacitor. A doped polysilicon strip is sometimes laminated with a metal silicide strip, and the doped polysilicon strip and the metal silicide strip form in combination the gate electrode, the bit line and the electrode of the stacked capacitor.

Various processes have been proposed to form a doped polysilicon layer. A typical example is a low-pressure chemical vapor deposition followed by a solid-phase diffusion. Intentionally undoped polysilicon is deposited through the low-pressure chemical vapor deposition, and the intentionally undoped polysilicon is treated with heat in an impurity source such as phosphorous trichloride. Then, the dopant impurity is diffused into the intentionally undoped polysilicon, and a doped polysilicon is obtained. The solid-phase diffusion may be replaced with an ion-implantation.

The doped silicon may be deposited through an in-situ doping during the low-pressure chemical vapor deposition. However, if the doped polysilicon is deposited over 600 degrees in centigrade, dopant concentration and the thickness are not uniform over the polysilicon layer, and, for this reason, it is recommendable to deposit doped amorphous silicon over 600 degrees in centigrade. The doped amorphous silicon is crystallized over 800 degrees in centigrade, and the crystallization imparts good conductivity to the doped silicon layer. Thus, the deposition technology for doped polysilicon is categorized into two groups, i.e., introducing dopant impurity into intentionally undoped polysilicon and the in-situ doping during the growth of silicon.

Comparing the two deposition technologies, the in-situ doping is simpler than the other, and is desirable, because the manufacturer wants to simplify a complicated process sequence for fabricating an ultra large scale integration.

FIGS. 1A to 1E illustrate a prior art process for fabricating a dynamic random access memory cell. The process starts with preparation of a p-type silicon substrate 1. A thick field oxide layer 1a is selectively grown on the major surface of the p-type silicon substrate 1, and defines an active area 1b assigned to a pair of dynamic random access memory cells.

The surface portion of the active area 1b is thermally oxidized at 700 degrees to 900 degrees in centigrade, and the active area 1b is covered with a thin gate oxide layer 2a. Phosphorous-doped amorphous silicon is deposited to 100 nanometers thick over the entire surface of the structure, and the phosphorous is introduced into the amorphous silicon at 1.5.times.10.sup.20 cm.sup.-3 during the growth. The phosphorous-doped amorphous silicon thus deposited over the entire surface forms a phosphorous-doped amorphous silicon layer 2b. Tungsten silicide is further deposited to 100 nanometers thick over the phosphorous-doped amorphous silicon layer 2b by using a sputtering technique, and forms a tungsten silicide layer 2c as shown in FIG. 1A. After the growth of the tungsten silicide layer 2c, phosphorous may be ion implanted into the phosphorous-doped amorphous silicon layer 2b at dose of 5.times.10.sup.19 cm.sup.-2 under acceleration energy of 30 KeV so as to compensate the phosphorous diffused from the phosphorous-doped amorphous silicon layer 2b into the tungsten silicide layer 2c. The phosphorous reaches the boundary between the phosphorous-doped amorphous silicon layer 2b and the tungsten silicide layer 2c.

Subsequently, a photo-resist etching mask (not shown) is patterned on the tungsten silicide layer 2c by using lithographic techniques, and the tungsten silicide layer 2c and the phosphorous-doped amorphous silicon layer 2b are selectively etched away through a dry etching technique. Thus, the phosphorous-doped amorphous silicon layer 2b and the tungsten silicide layer 2c are patterned into word lines WL, and the word lines WL is implemented by the lamination of the phosphorous-doped amorphous silicon strip 2da and the tungsten silicide strips 2db. Parts of word lines WL on the gate oxide layer 2a serve as gate electrodes 2d. The photo-resist etching mask is stripped off.

Using the gate electrodes as an ion-implantation mask, phosphorous or arsenic is ion implanted into the active area 1b, and forms an n-type source region 2e and an n-type drain region 2f as shown in FIG. 1B. The gate oxide layer 2a, the gate electrode 2d, the n-type source region 2e and the n-type drain region 2f as a whole constitute an n-channel enhancement type switching transistor 2.

Subsequently, boro-phosphosilicate glass is deposited to 500 nanometers thick over the n-channel enhancement type switching transistor 2, and the boro-phosphosilicate glass layer is reflowed in nitrogen ambience so as to improve the step coverage. The boro-phosphosilicate glass layer forms an inter-level insulating layer 3a.

While the boro-phosphosilicate glass layer is being reflowed, the phosphorous-doped amorphous silicon strips 2da are crystallized, and are converted to phosphorous-doped polysilicon strips 2dc. Thus, the word lines WL and, accordingly, the gate electrodes 2d are implemented by the lamination of the phosphorous-doped polysilicon strips 2dc and the tungsten silicide strips 2db.

A photo-resist etching mask (not shown) is provided on the inter-level insulating layer 3a, and has an opening of 350 nanometers square over the n-type drain region 2f. The photo-resist etching mask exposes a part of the inter-level insulating layer 3a to etchant used in a dry etching, and the etchant forms a bit contact hole 3b. The n-type drain region 2f is exposed to the bit contact hole 3b.

Phosphorous-doped amorphous silicon is deposited to 100 nanometers thick over the entire surface of the resultant structure, and the phosphorous concentration is of the order of 2.times.10.sup.20 cm.sup.-3. The phosphorous-doped amorphous silicon topographically extends the inner and upper surfaces of the inter-level insulating layer 3a, and forms a phosphorous-doped amorphous silicon layer. Tungsten silicide is topographically deposited to 100 nanometers thick over the entire surface of the phosphorous-doped amorphous silicon layer, and phosphorous is ion implanted into the boundary between the phosphorous-doped amorphous silicon layer and the tungsten silicide layer under the same conditions as the lamination of the amorphous silicon layer and the tungsten silicide layer on the gate oxide layer 2a.

A photo-resist etching mask (not shown) is provided on the tungsten silicide layer, and the lamination of the phosphorous-doped amorphous silicon layer and the tungsten silicide layer is respectively patterned into a phosphorous-doped amorphous silicon strip 4a and a tungsten silicide strip 4b. The phosphorous-doped amorphous silicon strip 4a and the tungsten silicide strip 4b form in combination a bit line BL as shown in FIG. 1C.

Phosphosilicate glass is deposited over the bit line BL on the first inter-level insulating layer 3a, and the phosphosilicate glass is reflowed. The phosphosilicate glass layer forms a second inter-level insulating layer 3c. While the phosphosilicate glass layer is being reflowed, the phosphorous-doped amorphous silicon strip 4a is crystallized, and is converted to a phosphorous-doped polysilicon strip 4c.

A photo-resist etching mask (not shown) is provided on the second inter-level insulating layer 3c, and exposes a part of the phosphosilicate glass layer to an etchant. The etchant selectively removes the phosphosilicate glass and the boro-phosphosilicate glass from the second inter-level insulating layer 3c and the first inter-level insulating layer 3a, and forms a node contact hole 3d. The n-type source region 2e is exposed to the node contact hole 3d.

Doped amorphous silicon is deposited over the entire surface of the resultant structure. The doped amorphous silicon fills the node contact hole 3d, and swells into a doped amorphous silicon layer 5a as shown in FIG. 1D.

A photo-resist etching mask (not shown) is provided on the doped amorphous silicon layer 5a, and exposes a part of the doped amorphous silicon layer 5a to an etchant. The etchant patterns the doped polysilicon layer 5a into a storage node electrode 5b.

The storage node electrode 5b is covered with a dielectric film structure 5c, and the doped amorphous silicon for the storage node electrode 5b is converted to a doped polysilicon during a heat treatment for the dielectric film structure 5c. Finally, the dielectric film structure 5c is covered with a cell plate electrode 5d, and the storage node electrode 5b, the dielectric film structure 5c and the cell plate electrode 5d form in combination a stacked storage capacitor 5 as shown in FIG. 1E.

Thus, the prior art process is simplified by using the in-situ doping. However, the prior art process encounters a problem in that the dopant impurity is diffused from the doped polysilicon strips 2dc and 4c. The diffusion results in a final dopant concentration lower than a target value and undesirable segregation of the dopant impurity.

As described hereinbefore, the layers of the semiconductor structure are repeatedly applied with heat. The heat diffuses the dopant impurity from the doped silicon strips into the tungsten silicide strips 2db/4b, and the dopant impurity is segregated at the boundaries between the doped polysilicon strip 2dc and the gate oxide layer 2a. The tungsten silicide has a large solubility. When the dopant concentration of a doped silicon strip is low, most of the dopant impurity is diffused into a tungsten silicide strip, and a depletion layer extends from the boundary between the doped silicon strip and the gate oxide layer during an operation. This results in deterioration of the transistor characteristics.

The ion-implantation after the deposition of the tungsten silicide compensates the dopant impurity diffused into the tungsten silicide layer. The ion-implantation after the deposition is hereinbelow referred to as "post ion-implantation". However, the ion implantation damages the crystal structure of the doped polysilicon, and a dispersion of dopant concentration takes place in the direction of depth of the doped polysilicon layer. A part of the ion-implanted impurity reaches the gate oxide layer 2a, and damages the gate oxide layer 2a.

A miniature field effect transistor for an ultra large scale integration has an extremely thin gate electrode, because a thick gate electrode deteriorates the step coverage over the miniature field effect transistor. In this situation, the post ion-implantation is liable to damage the gate oxide layer, and compensation is getting harder and harder.

If a doped polysilicon strip is initially doped with a large amount of dopant impurity, the doped polysilicon strip may be still heavily doped after the heat treatments. However, such a heavily doped amorphous silicon strip is very likely to segregate the dopant impurity at the boundary to the gate oxide layer. This is because of the fact that the crystallization of the amorphous silicon proceeds at a low speed under 850 degrees in centigrade. The dopant impurity is traveled over a long distance during the combination of silicon atoms, and is pushed out from the silicon layer. A low-temperature process becomes majority for an ultra large scale integration, because the ultra large scale integration requires shallow p-n junctions. The low-temperature process promotes "push-out" from the silicon layer. For this reason, the large initial dopant concentration does not eliminate the problem from the prior art process.

FIGS. 2A and 2B illustrate two impurity profiles PR1 and PR2 respectively created in two phosphorous-doped amorphous silicon layers. The first impurity profile PR1 and the second impurity profile PR2 indicate a relatively low dopant concentration and a relatively high dopant concentration, respectively, and the dopant concentrations represented by the impurity profiles PR1 and PR2 are 8.times.10.sup.19 cm.sup.-3 and 3.times.10.sup.20 cm.sup.-3. When the phosphorous-doped amorphous silicon layers are treated with heat, the phosphorous is diffused from the amorphous silicon close to the tungsten silicide layer toward the gate oxide layer, and the impurity profiles PR1/PR2 incline as indicated by plots PR1' and PR2' (see FIGS. 3A and 3B). The impurity profiles PR1'/PR2' teach us that the segregation takes place around the boundaries between the gate oxide layers and the doped silicon layers. Although the phosphorus is decreased from both boundaries between the doped silicon layers and the tungsten silicide layers, the decrement of the doped silicon layer with the relative low dopant concentration is drastic, and the segregated phosphorous is recrystallized during a heat treatment carried out thereafter. The recrystallized phosphorous imparts a large stress to the gate oxide layer, and decreases the withstanding voltage of the gate oxide layer.

FIG. 4 illustrates the relative frequency of damaged gate oxide layer in terms of the strength of electric field. When the leakage current density reaches 0.1 mA/cm.sup.2, the gate oxide layer is evaluated to be damaged. The B-mode defect is 10 to 20 percent around 4 to 8 MV/cm.

The segregation of dopant impurity is observed in a contact hole. As described hereinbefore, the bit line BL is held in contact with the n-type drain region 2f, and the storage node electrode 5b is held in contact with the n-type source region 2e. In the contact holes 3b/3d, the dopant impurity is also pushed out during the crystallization of the doped amorphous silicon strips 4a/5b, and tends to be segregated at the boundaries between the doped silicon strips 4a/5b and the n-type source/drain regions 2e/2f. The segregated dopant impurity is further diffused into the silicon substrate 1 during a heat treatment. Upon completion of the contact hole 3b/3d, the semiconductor structure is taken out from the etching system, and is moved to a deposition system. While an operator is conveying the semiconductor structure from the etching system to the deposition system, the n-type source/drain region 2e/2f is unavoidably exposed to the atmosphere, and a natural oxide is grown on the n-type source/drain region 2e/2f. The phosphorous is segregated into the natural oxide, and the phosphorous-doped natural oxide serves as a diffusion source. FIG. 5 illustrates an impurity profile PR3 around the contact between the n-type source/drain region 2e/2f and the doped polysilicon strip 4c/5b. The impurity profile PR3 indicates the segregation around the boundary between the n-type source/drain region 2e/2f and the doped polysilicon strip 4c/5b.

The diffusion source supplies the phosphorous through the n-type source/drain region 2e/2f into the p-type silicon substrate 1. The phosphorous thus diffused into the silicon substrate 1 changes the channel length, and decreases the threshold of a parasitic transistor.

Thus, the dopant impurity diffused from the doped silicon layer is causative of the deterioration of the circuit components of the prior art dynamic random access memory device.

It is therefore an important object of the present invention to provide a process of fabricating a semiconductor device which restricts a diffusion of dopant impurity from a doped silicon layer and a segregation of the dopant impurity.

To accomplish the object, the present invention proposes to form an intentionally undoped amorphous semiconductor material under a doped amorphous semiconductor material.

In accordance with the present invention, there is provided a process of fabricating a semiconductor device, comprising the steps of: a) forming an insulating layer; b) forming an intentionally undoped amorphous silicon layer on the insulating layer; c) forming a doped amorphous silicon layer containing a dopant impurity over the intentionally undoped amorphous silicon layer; and d) heating the intentionally undoped amorphous silicon layer and the doped amorphous silicon layer so as to crystallize the intentionally undoped amorphous silicon layer and the doped amorphous silicon layer.

The features and advantages of the process according to the present invention will be more clearly understood from the following description taken in conjunction with the accompanying drawings in which:

FIGS. 1A to 1E are cross sectional views showing the prior art process of fabricating the dynamic random access memory cell;

FIGS. 2A and 2B are graphs showing the impurity profiles before the heat treatment;

FIGS. 3A and 3B are graphs showing the impurity profiles after the heat treatment;

FIG. 4 in a graph showing the relative frequency of damaged gate oxide layers in terms of the strength of electric field;

FIG. 5 is a graph showing the impurity profile around the contact between the impurity region and the doped polysilicon strip;

FIGS. 6A to 6F are cross sectional views showing the process of fabricating a semiconductor dynamic random access memory device according to the present invention;

FIG. 7 is a graph showing the gas glow rate in terms of time during a deposition of amorphous silicon for a gate electrode;

FIG. 8 is a graph showing the phosphorous content varied with the thickness of the amorphous silicon layer;

FIG. 9 is a graph showing an impurity profile created in an amorphous silicon layer;

FIG. 10 is a graph showing an impurity profile created in a polysilicon layer;

FIG. 11 is a graph showing a relative frequency of damaged gate oxide layer in terms of the strength of electric field;

FIG. 12 is a cross sectional view showing the structure of a semiconductor dynamic random access memory device upon growth of an amorphous silicon layer;

FIG. 13 is a graph showing gas flow rates during a deposition of the amorphous silicon layer;

FIG. 14 is a graph showing a phosphorous concentration in the amorphous silicon layer;

FIG. 15 is a graph showing an impurity profile in an amorphous silicon layer before a growth of an inter-level insulating layer;

FIG. 16 is a graph showing an impurity profile in the amorphous silicon layer upon completion of a semiconductor dynamic random access memory device;

FIG. 17 is a graph showing a relative frequency of damaged gate oxide layer in terms of the strength of electric field;

FIGS. 18A and 18B are cross sectional views showing essential steps of yet another process according to the present invention;

FIG. 19 is a graph showing flow rates during a deposition of an amorphous silicon layer;

FIG. 20 is a graph showing a phosphorous concentration in the amorphous silicon layer;

FIG. 21 is a graph showing an impurity profile across the boundary between a silicon substrate and a polysilicon after a heat treatment;

FIG. 22 is a graph showing flow rates of process gas for depositing an amorphous silicon layer; and

FIG. 23 is a graph showing an impurity profile of the amorphous silicon layer.

Referring to FIGS. 6A to 6F of the drawings, a process of fabricating a semiconductor dynamic random access memory device embodying the present invention starts with preparation of a p-type silicon substrate 11. A thick field oxide layer 11a is selectively grown on the major surface of the p-type silicon substrate 11, and defines an active area 11b assigned to a pair of dynamic random access memory cells.

The surface portion of the active area 11b is thermally oxidized at 700 degrees to 900 degrees in centigrade, and the active area 11b in covered with a thin gate oxide layer 12a of 10 nanometers thick.

Subsequently, the semiconductor structure, i.e., the p-type silicon substrate 11 covered with the thick field oxide layer 11a and the thin gate oxide layer 12a is placed in a reactor (not shown) of a low-pressure chemical vapor deposition system, and amorphous silicon layer 12b is deposited to 100 nanometers thick over the entire surface of the semiconductor structure. First, only silane is introduced into the reactor (see FIG. 7), and intentionally undoped amorphous silicon 12c is deposited for 3 to 4 minutes. An intentionally undoped amorphous silicon layer 12c is formed on the thick field oxide layer 11a and the thin gate oxide layer 12a as shown in FIG. 6A, and is of the order of 10 nanometers thick. The gas pressure is regulated to 2 torr, and silane is decomposed at 530 degrees in centigrade.

Subsequently, phosphine is mixed into the silane (see FIG. 7), and the gas pressure and the temperature is maintained. The phosphine is mixed into nitrogen at 1 percent, and the gaseous mixture of phosphine and nitrogen is hereinbelow simply referred to as "phosphine gas". The phosphine gas is increased at 5 sccm/min. When the phosphine gas reaches 20 sccm, the phosphine gas is maintained at 20 sccm for 20 minutes. Then, a lightly phosphorous-doped amorphous silicon layer is deposited to 50 nanometers thick over the intentionally undoped amorphous silicon layer 12c, and the lightly phosphorous-doped amorphous silicon layer contains the phosphorous at 0.8.times.10.sup.20 cm.sup.-3. The phosphine gas is increased to 100 sccm at 10 sccm/minute, and the silane gas is decreased to 800 sccm at 40 sccm/minute. A heavily phosphorous-doped amorphous silicon layer is deposited to 40 nanometers thick for 10 minutes, and the phosphorous concentration is of the order of 3.times.10.sup.20 cm.sup.-3. Thus, the phosphorous-doped amorphous silicon layer 12d of 90 nanometers thick is formed on the intentionally undoped amorphous silicon layer 12c, and the intentionally undoped amorphous silicon layer 12c and the phosphorous-doped amorphous silicon layer 12d form the amorphous silicon layer 12b of 100 nanometers thick. The deposition speed under the above described conditions is substantially constant regardless of the phosphine, and ranges from 2.4 nanometers to 2.8 nanometers per minute. The phosphorous content in the amorphous silicon layer 12b is represented by plots PR4 in FIG. 8.

Subsequently, tungsten silicide is further deposited to 100 nanometers thick over the amorphous silicon layer 12b by using a sputtering technique, and forms a tungsten silicide layer 12e as shown in FIG. 6B. No phosphorous is ion implanted into the boundary between the tungsten silicide layer 12e and the heavily phosphorous doped amorphous silicon layer 12d.

Subsequently, a photo-resist etching mask (not shown) is patterned on the tungsten silicide layer 12e by using lithographic techniques, and the tungsten silicide layer 12e and the amorphous silicon layer 12b are selectively etched away through a dry etching technique. Thus, the amorphous silicon layer 12b and the tungsten silicide layer 12e are patterned into word lines WL, and the word lines WL is implemented by the lamination of an intentionally undoped amorphous silicon strip 12da, a phosphorous-doped amorphous silicon strip 12db and a tungsten silicide strip 12dc. A parts of word line WL on the gate oxide layer 12a serves as a gate electrode 12d. Upon completion of the patterning, the photo-resist etching mask is stripped off.

Using the gate electrodes 12d as an ion-implantation mask, phosphorous or arsenic is ion implanted into the active area 11b, and forms an n-type source region 12e and an n-type drain region 12f as shown in FIG. 6C. The gate oxide layer 12a, the gate electrode 12d, the n-type source region 12e and the n-type drain region 12f as a whole constitute an n-channel enhancement type switching transistor 12.

Subsequently, boro-phosphosilicate glass is deposited to 500 nanometers thick over the n-channel enhancement type switching transistor 12, and the boro-phosphosilicate glass layer is reflowed around 850 degrees in centigrade in nitrogen ambience so as to improve the step coverage. The boro-phosphosilicate glass layer forms an inter-level insulating layer 13a.

While the boro-phosphosilicate glass layer is being reflowed, the amorphous silicon strips 12b are crystallized, and are converted to polysilicon strips 12b'. Thus, the polysilicon strip 12b' and the tungsten silicide strip 12dc form in combination the word line WL and, accordingly, the gate electrode 12d.

A photo-resist etching mask (not shown) is provided on the inter-level insulating layer 13a, and has an opening of 350 nanometers square over the n-type drain region 12f. The photo-resist etching mask exposes a part of the inter-level insulating layer 13a to etchant used in a dry etching, and the etchant forms a bit contact hole 13b. The n-type drain region 12f is exposed to the bit contact hole 13b.

Conductive material or materials are deposited over the entire surface of the resultant semiconductor structure, and swells into a conductive layer. A photo-resist etching mask (not shown) is provided on the conductive layer, and the conductive layer is patterned into a bit line BL as shown in FIG. 6D.

Phosphosilicate glass is deposited over the bit line BL on the first inter-level insulating layer 13a, and the phosphosilicate glass is reflowed. The silicon for the word lines WL is further crystallized during the heat treatment. The phosphosilicate glass layer forms a second inter-level insulating layer 13c. A photo-resist etching mask (not shown) is provided on the second inter-level insulating layer 13c, and exposes a part of the phosphosilicate glass layer to an etchant. The etchant selectively removes the phosphosilicate glass and the boro-phosphosilicate glass from the second inter-level insulating layer 13c and the first inter-level insulating layer 13a, and forms a node contact hole 13d. The n-type source region 12e is exposed to the node contact hole 13d.

Conductive material is deposited over the entire surface of the resultant structure. The conductive material fills the node contact hole 13d, and swells into a conductive layer 15a as shown in FIG. 6E.

A photo-resist etching mask (not shown) is provided on the conductive layer 15a, and exposes a part of the conductive layer 15a to an etchant. The etchant patterns the conductive layer 15a into a storage node electrode 15b.

The storage node electrode 15b is covered with a dielectric film structure 15c, and the polysilicon layer 12b' is further crystallized during a heat treatment for the dielectric film structure 15c. The heat treatment is carried out at 850 degrees in centigrade. Finally, the dielectric film structure 15c is covered with a cell plate electrode 15d, and the storage node electrode 15b, the dielectric film structure 15c and the cell plate electrode 15d form in combination a stacked storage capacitor 15 as shown in FIG. 6F.

The present inventor measured the dopant As concentration in the amorphous silicon layer 12b before the growth of the inter-level insulating layer 13a and the dopant concentration in the polysilicon layer 12b' upon completion of the semiconductor dynamic random access memory device. The present inventor plotted the dopant concentration in the amorphous silicon layer 12b as indicated by plots PR5 in FIG. 9 and the dopant concentration in the polysilicon layer 12b' as indicated by plots PR6 in FIG. 10. The impurity profile represented by the plots PR5 stepwise increased from the boundary between the gate oxide layer 12a and the amorphous silicon layer 12b. However, the impurity profile PR6 became flat after the heat treatment. The dopant impurity around the boundary between the amorphous silicon layer 12b and the tungsten silicide layer 12e was so large that the dopant impurity was still large after the diffusion from the boundary toward the intentionally undoped amorphous silicon layer 12c. On the other hand, the intentionally undoped amorphous silicon layer 12c was contiguous to the gate oxide layer 12a, and the phosphorous was hardly segregated at the boundary therebetween.

The present inventor further evaluated the influence of segregated phosphorous. The present inventor applied an electric field across the gate oxide layers 12a, and increased the strength of electric field to see whether or not the large amount of leakage current flowed across the gate oxide layer 12a. When the leakage current exceeded 0.1 mA/cm.sup.2, the present inventor decided it to be damaged. The relative frequency of damaged gate oxide layers was plotted in FIG. 11. Comparing FIG. 11 with FIG. 4, it was understood that the damaged gate oxide layers 12a were negligible below 9.5 MV/cm, and this tendency taught us that the segregated phosphorous was little. Thus, the intentionally undoped amorphous silicon layer 12c improved the gate oxide layer 12a.

Another process embodying the present invention is described hereinbelow. The process implementing the second embodiment is similar to the first embodiment except for the growth of an amorphous silicon layer, and, for this reason, only the growing step is illustrated in FIG. 12.

The process also starts with preparation of a p-type silicon substrate 21, and a thick field oxide layer 21a is selectively grown on the major surface of the p-type silicon substrate 21. The thick field oxide layer 21a defines an active area assigned to two dynamic random access memory cells, and a surface portion of the active area is thermally oxidized. A thin gate oxide layer 22a is grown on the active area.

Subsequently, an amorphous silicon layer 23 is deposited to 100 nanometers thick by using the low pressure chemical vapor deposition system. The reactor of the low pressure chemical vapor deposition system is maintained at 1.8 torr, and the amorphous silicon is grown at 530 degrees in centigrade.

FIG. 13 illustrates gas flow rates during the deposition of the amorphous silicon layer 23. First, only silane gas is introduced into the reactor for 3 to 4 minutes, and intentionally undoped amorphous silicon layer 23a is deposited to 10 nanometers thick over the gate oxide layer 22a. The phosphine gas is mixed into the silane gas, and is increased to 20 sccm at 5 sccm/min, and a first lightly phosphorous doped amorphous silicon layer 23b is deposited over the intentionally undoped amorphous silicon layer 23a.

When the flow rate of the phosphine gas reaches 20 sccm, the silane gas and the phosphine gas are stopped, and oxygen diluted in nitrogen at 1 percent is introduced into the reactor at 15 sccm, and the resultant semiconductor structure is exposed to the oxygen gas for 5 minutes. The oxygen is absorbed in a surface portion of the first lightly phosphorous doped amorphous silicon layer 23b, and a first absorbed layer 23c of 1 nanometer thick is formed on the first lightly phosphorous doped amorphous silicon layer 23b.

Subsequently, the silane gas and the phosphine gas are supplied to the reactor again at 1200 sccm and 20 sccm, respectively, and a second lightly phosphorous doped amorphous silicon layer 23d is grown for 20 minutes.

The silane gas and the phosphine gas are stopped again, and the oxygen gas is introduced into the reactor at 15 sccm for 5 minutes. A second absorbed layer 23e is formed on the second lightly phosphorous doped amorphous silicon layer 23d.

Finally, the silane gas and the phosphine gas are introduced into the reactor at 800 sccm and 100 sccm, and a heavily phosphorous doped amorphous silicon layer 23f is grown on the second absorbed layer 23e for 15 minutes. The phosphorous concentration is varied as shown in FIG. 14.

Upon completion of the growth of amorphous silicon, tungsten silicide is deposited to 100 nanometers thick over the heavily phosphorous doped amorphous silicon layer 23f by using a sputtering, and a tungsten silicide layer 24 is laminated on the amorphous silicon layer 23.

A photo-resist etching mask (not shown) is prepared on the tungsten silicide layer 24, and the amorphous silicon layer 23 and the tungsten silicide layer 24 are patterned into a word line. A part of the word line on the gate oxide layer 22a serves as a gate electrode.

An insulating material is deposited over the word lines, and is reflowed. While heat is reflowing the insulating layer, the amorphous silicon layer 23 is crystallized, and is converted to a polysilicon layer. The insulating layer serves as a lower inter-level insulating layer.

A bit contact hole is formed in the lower inter-level insulating layer, and conductive material is deposited over the lower inter-level insulating layer. The conductive material fills the bit contact hole, and swells into a conductive layer. The conductive layer is patterned into bit lines.

An upper inter-level insulating layer is formed on the bit lines, and the crystallization proceeds during the reflow of the upper inter-level insulating layer. A node contact hole is formed in the upper inter-level insulating layer and the lower inter-level insulating layer, and conductive material is deposited over the upper inter-level insulating layer. The conductive layer is patterned into a storage node electrode, and the storage node electrode is covered with a dielectric film which in turn is covered with a cell plate electrode. The crystallization further proceeds during a heat treatment for the dielectric film.

The present inventor measured the phosphorous concentration in the amorphous layer 23 and in the polysilicon layer, and plotted the phosphorous concentration in FIGS. 15 and 16. Upon completion of the growth of the amorphous silicon layer 23, the phosphorous concentration was represented by plots PR7, and was decreased to zero at the first and second absorbed layers 23c/23e. Upon completion of the semiconductor dynamic random access memory device, the phosphorous was segregated in the first and second absorbed layers 23c/23e, and the segregated phosphorous was drastically decreased at the boundary between the gate oxide layer 22a and the polysilicon layer. The present inventor thought that the first and second absorbed layers 23c/23e trapped the phosphorous diffused from the heavily phosphorous doped amorphous silicon layer 23f.

The present inventor further evaluated the influence of the segregation. The present inventor increased the strength of electric field across the gate oxide layers, and measured leakage current flowing across the gate oxide layers. The relative frequency of damaged gate oxide layers was plotted in FIG. 17, and confirmed that the first and second absorbed layers 23c/23e were effective against the segregation of phosphorous.

The present inventor further confirmed that, although the first and second absorbed layers 23c/23e slightly increased the resistance of the gate electrode, the increment was negligible and no parasitic capacitor took place in the gate electrode. The first and second absorbed layers 23c/23e range from 0.5 nanometer thick to 2.0 nanometers thick.

Yet another process embodying the present invention was described hereinbelow. The process implementing the third embodiment is similar to that of the first embodiment until the formation of the node contact hole 13d. For this reason, regions and layers of the semiconductor structure shown in FIGS. 18A and 18B are labeled with references designating corresponding regions and layers of the first embodiment without detailed description.

After the formation of the node contact hole 13d, the resultant semiconductor structure is placed in a reactor (not shown) of a low pressure chemical vapor deposition system, and an amorphous silicon layer 31 of 300 nanometers thick is deposited over the entire surface of the second inter-level insulating layer 13c.

FIG. 19 illustrates gas flow rates during the chemical vapor deposition. First, silane gas is introduced into the reactor at 1200 sccm for 5 minutes, and an intentionally undoped amorphous silicon layer 31a is topographically grown to 10 to 15 nanometers thick. The intentionally undoped amorphous silicon layer 31a defines a secondary node contact hole in the node contact hole 13d. Thereafter, the phosphine gas is gradually increased to 20 sccm, and is maintained at 20 sccm for 2 hours. Phosphorous-doped amorphous silicon fills the secondary node contact hole, and swells into a phosphorous-doped amorphous silicon layer 31b over the second inter-level insulating layer 13c as shown in FIG. 18A. The total thickness of the intentionally undoped amorphous silicon layer 31a and the phosphorous-doped amorphous silicon layer 31b is 300 nanometers. The dopant impurity is represented by plots PR9 of FIG. 20.

The resultant semiconductor structure is taken out from the reactor of the low-pressure chemical vapor deposition system, and is conveyed to a spin coater (not shown), by way of example. A photo-resist etching mask (not shown) is formed on the phosphorous-doped amorphous silicon layer 31b, and the amorphous silicon layer 31 is patterned into a storage node electrode 32. The storage node electrode 32 is implemented by the lamination of an intentionally undoped amorphous silicon strip and a phosphorous-doped amorphous silicon strip.

The resultant semiconductor structure is placed in the reactor of the low-pressure chemical vapor deposition system, and gaseous mixture of dichlorosilane and ammonia is introduced into the reactor. The gaseous mixture is maintained at 0.3 torr, and silicon nitride is deposited to 5 nanometers thick over the entire surface of the resultant semiconductor structure at 800 degrees in centigrade. While the silicon nitride is being deposited, the amorphous silicon layer 31 is crystallized, and the phosphorous is diffused from the phosphorous-doped amorphous/polysilicon strip toward the intentionally undoped amorphous/polysilicon strip. The storage node electrodes 32c is implemented by the lamination of polysilicon strips 32a/32b, and are covered with dielectric films 32d.

Finally, doped polysilicon is deposited to 150 nanometers thick over the dielectric films 32d, and a doped polysilicon layer is patterned into a cell plate electrode 32e as shown in FIG. 18B.

The present inventor measured the phosphorous concentration in the storage node electrode 32c upon completion of the semiconductor dynamic random access memory device, and the phosphorous concentration was plotted in FIG. 21. Comparing the impurity profile PR10 of FIG. 21 with the impurity profile PR3 of FIG. 5, it was understood that the intentionally undoped amorphous silicon layer effectively prevented the boundary with the silicon substrate 11 from segregation of phosphorous. Moreover, the diffusion range from the boundary was shorter than that of the prior art. Thus, the intentionally undoped amorphous silicon layer 31a restricts the out diffusion of phosphorous, and does not deteriorate the dynamic random access memory cell.

Absorbed layers may be formed in the phosphorous doped amorphous silicon layer 31b. In order to form the absorbed layers, the silane gas and the phosphine gas are intermittently stopped, and the oxygen gas is introduced into the reactor as shown in FIG. 22. Then, one or two absorbed layers are formed in the phosphorous-doped amorphous silicon layer 31b around 10 to 50 nanometers spaced from the boundary between the intentionally undoped amorphous silicon layer 31a and the phosphorous doped amorphous silicon layer 31b. The phosphorous concentration of the amorphous silicon layer 31 is shown in FIG. 23. The present inventor confirmed that the absorbed layers were effective against the diffusion of the phosphorous.

Although particular embodiments of the present invention have been shown and described, it will be obvious to those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the present invention.

For example, an intentionally undoped amorphous silicon strip, a lightly phosphorous doped amorphous silicon strip, a heavily phosphorous doped amorphous silicon strip and a refractory metal silicide strip may be converted to a lamination of a polysilicon strip and the refractory metal silicide strip during a reflow of an inter-level insulating layer.

The present invention is applicable to a bit line and other signal wirings.

The dopant impurity is never limited to phosphorous. A doped polysilicon may contain arsenic, antimony, boron or another p-type dopant impurity. The refractory metal silicide may be titanium silicide, molybdenum silicide etc.