JPS5856476A - Semiconductor pressure converter - Google Patents

Abstract

PURPOSE:To obtain a semiconductor pressure converter having sufficient bonding strength and high utility by bonding a pressure sensitivie pellet to its base via an adhesive layer made of solder glass. CONSTITUTION:An Nichiden glass LS-0110 is, for example, formed by a precipitation method in a thickness of 150mum on an Si base 18 to be bonded with a pressure sensitive element wafer 17. In the method, 5g of a solder glass is dissolved in 500cc of ethyl alcohol, a sample is placed at the position of 1cm under the water surface, and is precipitated for approx. 1hr. A bonding pressure sensitive wafer 17 is opposed to an Si base 18, a weight of approx. 10g/cm<3> is placed, and is bonded in a furnace at approx. 470 deg.C. The wafer after bonding is laterally and longitudinally cut at the positions of arrows, thereby obtaining pellets. The bonding with the package is performed on the back surface 14' of the Si base 14.

Description

DETAILED DESCRIPTION OF THE INVENTION It is well known that when a semiconductor single crystal is strained by stress, its resistance value changes. By effectively utilizing this physical phenomenon and combining it with semiconductor IC technology, compact and high-performance semiconductor pressure transducers (pressure sensors) have been developed in recent years.

FIG. 1 is a schematic diagram showing the structure of a semiconductor diffusion diaphragm turgor pressure sensor that has not yet been developed for high-precision industrial use. This sensor consists of 1Q made of n-8i (100) simple crystal.
In the center of pellet 1 of ss', about 4 to 5
It has a structure in which a strain diaphragm 2 processed to have a thickness of about 70 μm in diameter sφ is provided, and a piezoresistance gauge 3 is provided on the surface of the diaphragm 2 by thermally diffusing boron (Bl) into strips. 1 is fixed on the pedestal 4, mounted on the package 5, and wired to the electrode lead by metal @6.
A pipe 7 provided in communication with the through hole introduces fluid pressure into the diaphragm portion.

Therefore, the pedestal 4 can be viewed from the outside, especially the package 5.
This sensor functions as a strain barrier material that prevents strain from being transmitted to the diffused resistance strain gauge 3. Due to the structure that prevents such external strain, the sensor can be used in a wide temperature range of -30 to 85C. It has a high performance called 'lFS.

However, such high-precision industrial pressure sensors employ a special package structure as described above, and therefore are extremely expensive. In addition, not only the package but also the element part uses silicon material, but mass-produced ICs
The technology remains unutilized. Furthermore, in recent years, with the spread of microcomputers, there have been many requests for semiconductor pressure sensors, but in this case, lower costs are desired rather than higher performance.

Therefore, recently, a low-cost semiconductor pressure sensor having the structure shown in FIGS. 2(a) to 2(b) has been developed. This sensor consists of a pressure-sensitive diaphragm 1 part and a pedestal 4 shown in FIG.
By introducing a silicon diffusion process, the parts can be mass-produced. 4mo, 3QO according to Figure 2 (a)
A diaphragm 12 having a thickness of 2 Wtl 114 opm is formed from the back surface of the silicon pellet 11 having a thickness of μm, and a diffused resistance strain gauge 13 is provided on the surface of the J2 portion of the diaphragm. These are formed similarly to those shown in FIG. 1, and in the above example are sensitive to a fluid pressure of 21jf/DIII. However, pellet 1
1 is bonded by glass 15 onto a pedestal 14 made of silicon, and fluid pressure is introduced into the diaphragm 12 through a through hole 16 provided in the pedestal 14. This pedestal 14 prevents stress other than fluid pressure from being applied to the strain gauge 13 formed on the pellet 11 like the pedestal 4 described above, and is made as thick as possible. Similarly, generally, a pedestal 14 with a thickness of 1 m to 5 se is used due to limitations in the processing technology of the through hole 16. A pressure sensor with such a structure is shown in Fig. 2 (bl, (cl),
Pellets 11 with 4 ms openings were formed in a matrix on a silicon wafer 17, and a sea urchin/X fin base 18 with through holes 16 was bonded to the silicon wafer 17 via a glass adhesive layer 15. In this case, the pellet diffusion process is achieved through the semiconductor process, which is almost the same as that of conventional bipolar ICs. , and the shape of the diaphragm can also be achieved by etching techniques using alkaline chemicals.This greatly improves the manufacturing process of pressure sensors and makes great strides in changing their quantity.

However, the most important issue in order to facilitate such manufacturing is the bonding of the silicon wafer on which the pressure-sensitive element is provided and the wafer-shaped base with the through-hole. The key is whether or not it can be done without damaging its characteristics.

Wafer 11 for pressure sensitive element and through hole 1 in FIG. 2(b)
In general, glass adhesion and metal melt adhesion are considered as means for adhering the Si bases 18 having 8 to face each other.

In order to melt and bond silicon single crystals with metal, it is necessary to metalize the underlying silicon, which complicates the wafer processing process. Furthermore, the coefficient of thermal expansion of molten metal as an adhesive is generally approximately -0.0 digits larger than that of a silicon single crystal (31×10'), and causes a large adhesive strain to the silicon pressure-sensitive element. On the other hand, the former method of glass bonding does not require pretreatment such as metallization on the silicon wafer. In addition, solder glass is used for bonding pressure sensors because it is possible to select solder glass whose coefficient of thermal expansion is smaller than that of metal.

In this case, there are two major conditions that require the use of a laminated glass in order to bond the pressure-sensitive element on the wafer and the silicon base as shown in FIG.

One of these is that the chemical temperature at which the glass is melted must be 550C or less, preferably 5oOC or less, in order to prevent deterioration of the metal wiring layer (generally aluminum wiring S) of the pressure-sensitive element, and the heat of the adhesively hardened glass layer. Two points are that the coefficient of expansion is as close as possible to the coefficient of thermal expansion of silicon.

There is no commercially available or developed solder glass that completely satisfies the above conditions. Due to the characteristics of solder glass, the coefficient of thermal expansion of silicon (31xlO
') and the closer it is, the higher the working temperature is. First, Table 1 lists the working temperatures and coefficients of thermal expansion of commercially available solder glasses under conditions lower than the working temperature of 550C.

Table 1 When soldering the solder glass listed here, it is important to first prepare the adhesive layer of the solder glass as shown in FIGS. 2(a) and 2(b). Since the wafer 17 serving as the pressure-sensitive element and the 31 units 16 are bonded together, it is desirable that the adhesive layer 16 be as thin as possible. According to experiments, it was possible to form a thin layer using the precipitation method for all glasses with sample numbers "1, 2, 3." This Si base is placed and allowed to settle for about 1 hour.A glass adhesive layer with a thickness of about 30 μm is formed.Using the glass adhesive layer formed in this way, As a result of considering glass,
The results shown in Table 2 were obtained.

Table 2: With glass 2.3 where the working temperature is 500C or less, the HL wiring does not deteriorate, but there are many cracks in the glass, and 30-40% of the wires do not reach the guaranteed airtight limit of 10-' atm 6 CC/min. It comes out. It is said that it is desirable to make the glass layer as thin as possible to avoid the difference between the coefficient of thermal expansion of the glass layer and that of silicon. Various precipitation methods have been investigated and attempts have been made to make the glass layer thinner, but when the thickness is less than 20 μm, it becomes difficult to bond the entire surface of the glass layer (for example, to a 2-layer wafer).

On the other hand, cracks in the glass layer were still observed, and it was found that the leakage failure of the bonded sample reached 50%. In other words, if the expansion coefficients are different, the leakage problem cannot be solved by making the adhesive layer thinner. In terms of leakage alone, the yield is better if the adhesive layer has a certain thickness.

On the other hand, the glass shown in sample number 4.5 has a low working temperature, and its coefficient of thermal expansion is that of sample numbers 1, 2.
smaller than any of the 3. Although efforts were made to make full use of this glass, it was found that even if a precipitate layer of about 30 am was formed, it would not become glassy with adhesive strength. but,
It was found that at 1002 m or more, the glass becomes vitrified and the adhesive force increases, and that there is no crank at all in the glass layer coated on silicon for a length of 70 m or more. This is because the solder glass contains a lot of filler, and when the adhesive layer approaches the size of the filler particles, the solder glass does not fill in between the filler particles. It depends on not becoming a person.

Conventionally, when the coefficient of thermal expansion of the adhesive layer and the coefficient of thermal expansion of the adhered object are different, it has been thought that it is desirable to make the adhesive layer as thin as possible in order to avoid the influence of the difference. However, it has been found that when a large amount of filler is contained and the thermal expansion is reduced as a result, the adhesive layer is more effective in absorbing strain if the adhesive layer is made larger than a certain level.

The present invention has been made based on this paradoxical situation, and its purpose is to provide a highly practical semiconductor pressure transducer that has sufficient adhesive strength.

Examples of the present invention will be described below. For example, in FIG. 2, on the Bi stand 18 to which the pressure sensitive element wafer 11 is bonded,
For example, by precipitation method, 8den glass LS-0110 is 150 μm thick.
Form to a thickness of . In the precipitation method described above, 100C
The solder glass was dissolved in H ethyl alcohol, and the sample was placed 1 cm below the water surface and allowed to settle for about 1 hour. The "pressure sensitive wafer 11 to be bonded" is placed facing the Si stand 18, and a weight of about 10P/- (for a 2 o'clock wafer, about 2
As shown in Figure 2(C), the wafer after bonding is cut vertically and horizontally at the arrow positions to obtain pellets as shown in Figure 2(a). I can do it. Thus, the bonded pellets have a higher leakage yield than any of the Panda glasses of Sample Nos. 1, 2, and 3 mentioned above, and for example, those with a leakage rate of 10-' atm-CC/min or more in 100 pellets trial production. There were only 4 cases. In other words, the yield is 96
It was Chi. It was also found that other properties (electrical properties, temperature properties) were statistically superior to the solder glasses of sample numbers 1 and 2°3. The thus formed pellets are again bonded to the shell cage member for applying pressure, and the bonding method at this time can be either solder glass or epoxy bonding, depending on cost and usage. The bonding with the package is done on the back surface J4' of the Bi base J4 in FIG. None. That means the thickness is 7
It has been confirmed that the effect is lost when the pressure is less than 0 mm.As described above, the semiconductor pressure transducer according to the present invention has a pressure-sensitive pellet and its base with a thermal expansion coefficient of 40 to 70xlO'. ,
In addition, it has a structure in which it is bonded via an adhesive layer made of solder glass with a thickness of 70 pm or more, which reduces distortion to the pressure-sensitive pellet due to this adhesive layer and can sufficiently ensure its complete airtightness. . Moreover, the adhesive workability is easy, and the adhesive has many practical advantages.

In addition, #DreamInvention not only applies to this type of semiconductor pressure transducer, but also applies its technical philosophy to COD and LS with large pellet sizes.
It can also be applied to 91 elements for driving liquid crystals, etc.

[Brief explanation of the drawing]

Figure 1 is a configuration diagram of a high-precision industrial cow conductor pressure sensor, Figure 2
Figures (a) to (C1) are diagrams showing the structure of the element part of the individual pressure sensor and its manufacturing process. 1... Pellet, 2... Strain diaphragm, 3...
・Piezo resistance gauge, 4...91 base, 5... package, 6... bonding wire, 7... pressure introduction pipe, 11... pellet, 12... diaphragm, 13... piezo Resistance gauge, 14... Si base, 15... Glass adhesive layer, 16... Opening, 11.
...Pressure-sensitive triplet wafer, 18...B+base Eva. Applicant's agent Patent attorney Suzu Hoshiki Figure 1 (a)

Claims (1)

[Claims] A pressure-sensitive pellet is formed by forming a diffused resistance layer on the surface of a thin diaphragm formed in the center of a semiconductor substrate, and an opening for introducing pressure in the center of the pressure-sensitive pellet is placed facing the peripheral thick part of the pressure-sensitive pellet. a base made of a semiconductor material that blocks strain on the pressure-sensitive pellet;
1. A semiconductor pressure transducer comprising: an adhesive layer made of solder glass of 70xlO and a thickness of 70 μm or more.