JP2561280B2 - Photoelectric conversion device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention is applied to a line sensor of a facsimile, an image scanner, a digital copying machine, or the like, which reads image information while moving a document for image reading relatively. The present invention relates to a suitable photoelectric conversion device.

[Prior Art] Conventionally, as an image reading apparatus using a one-dimensional line sensor, a one-dimensional line sensor having a length of several cm is used to form a document image using a reduction optical system to read document image information. Things are known. However, this type of image reading apparatus requires a large optical path length for performing reduction or imaging, and it is difficult to make it compact because the volume of the optical system is large.

On the other hand, when an equal-magnification optical system using a long one-dimensional line sensor having the same length as the document width is used, the volume of the optical system can be remarkably reduced, and the reader can be downsized.

FIG. 8 is a schematic configuration diagram of an image reading apparatus using such a unit magnification optical system.

In the figure, the document P is transported in the f direction by a transport roller (not shown), and the sensor unit 50 provided on the lower surface of the document P reads the document image information. In this unit 50, a light source 51 as an illuminating device for illuminating the original P over the entire width of the original P, and an optical system 52 including a SELFOC lens array for forming reflected light from the original P at an equal magnification. A photoelectric conversion unit 53 that performs photoelectric conversion of reflected light is provided.

The photoelectric conversion unit 53 sequentially sends the signals photoelectrically converted by the optical sensor to the signal processing unit 55 with the clock signal from the control clock generation unit 54.

The signal processing unit 55 converts the analog signal photoelectrically converted by the conversion unit 53 into a digital signal by the AD converter 56, corrects the sensitivity variation by the correction table memory 57 and the correction data memory 58, and outputs the corrected signal. ing. Each of these units is controlled by a clock signal from the control clock generating unit 59.

9 and 10 are a schematic configuration diagram showing the configuration of the photoelectric conversion unit 53 and its equivalent circuit diagram.

In both figures, on a substrate 60, an optical sensor section 61 composed of optical sensors S _{1 to} S _n , a storage capacitor section 62 composed of capacitors C _{1 to} C _n , transfer switches T _{1 to} T _n and a flip-flop.
Transfer units 63 each including F _{1 to} F _n are formed. Note that the flip-flops F _{1 to} F _n form a shift register. The bias voltage V _S is applied to the optical sensors S _{1 to} S _n , and the clock pulse CP is supplied from the control clock generation unit 54 to the clock terminals of the flip-flops F _{1 to} F _n .

With such a configuration, the image information of the document is detected by the optical sensor.
The photocurrents are converted into photocurrents by S _{1 to} S _n , and the photocurrents are stored as storage charges in the storage capacitors C _{1 to} C _n for a certain period of time and output via the transfer unit 63.

[Problems to be Solved by the Invention] However, in such a conventional example, when a normal silicon chip is used as the driving IC chip that constitutes the transfer unit, the number of bonding connections between the optical sensor and this IC chip is large. Therefore, the cost increases and the yield decreases.

That is, for example, an A4 size original is 40
When reading with 0 pixels, the number of elements of the optical sensor is about 3,500. Therefore, the number of processing bits per IC chip is 6
With 4 bits, 54 IC chips are required, and the number of connection bondings is very large, 3500 to 5000.

For this reason, conventionally, a method has been adopted in which the transfer portion is formed of a thin film transistor using polycrystalline or amorphous silicon and is manufactured on the same substrate as the optical sensor. However, since the electron mobility of polycrystalline and amorphous silicon is as small as 1/100 to 1/1000 of that of crystalline silicon, the reading speed is limited, and it takes 15 to 30 seconds to read an A4 size original. Therefore, when used for a facsimile, there is a problem that it can be applied only to low-speed machines.

Further, since the variation in the sensitivity of the optical sensor is a characteristic peculiar to each photoelectric conversion unit, it is necessary to write the correction data in the correction data memory in advance before the document reading operation. further,
This correction data memory is a read / write memory (RW
M), it requires a backup with a battery to keep the contents of the memory even when the power of the device is turned off.

Therefore, it is possible to provide the correction data by using the read-only memory (ROM) as the correction data memory, but in this case, the variation in the sensitivity of each optical sensor is measured for each photoelectric conversion unit, and the It is necessary to write data to the ROM and attach this ROM to each conversion unit and incorporate it into the device, so parts management and transportation management in the assembly process of the device become very complicated, and the combination of the conversion unit and ROM May be mistaken and is not preferable in terms of reliability.

Further, different control clocks are required for reading the accumulated charge from the conversion unit and reading the correction data from the signal processing unit. This is to drive the n-stage shift register controlled by the shift pulse by the converter, and to drive a special power supply system that matches the characteristics of the optical sensor and the transfer switch. On the other hand, drive the signal processor. This is because the drive is for controlling the address of the memory. For this reason, two types of control clock generators are required, which increases the size and cost.

Due to the above problems, an image reading apparatus suitable for a high-performance line sensor, which is small in size and low in cost and which can simplify production management and processes, has not been obtained.

[Means for Solving the Problems] The photoelectric conversion device according to the present invention has a non-nucleation surface having a small nucleation density and an area sufficiently small for crystal growth to form nuclei of only a single nucleus. An optical sensor section on a semiconductor single crystal grown from the single nucleus by subjecting a substrate having a free surface disposed adjacent to a nucleation surface having a nucleation density higher than the density to a crystal growth treatment, A transfer switch section for outputting signals of the optical sensor section in time series, a drive section for driving the transfer switch section, and a correction memory section for storing information on variation in sensitivity of each optical sensor element of the optical sensor section are formed. It is characterized by having done.

[Operation] As described above, by using the technique of forming the nucleation surface and the non-nucleation surface on the substrate and selectively growing the semiconductor single crystal, a transfer unit such as an optical sensor, a transfer switch, a shift register, or the like on the substrate, By making a correction memory, etc., a high-speed, high SN ratio, small-sized, low-cost device is possible.

Example First, a technique for selectively growing a semiconductor single crystal on an insulating substrate used in the present invention will be described.

FIGS. 11A and 11B are explanatory views of the selective nucleation crystal growth method.

The selective nucleation crystal growth method uses the difference between materials such as surface energy, sticking coefficient, desorption coefficient, surface diffusion rate, and other factors that influence nucleation in the thin film formation process. In this method, a single crystal is selectively caused to occur at a desired position of, and a crystalline deposited film is formed based on this nucleus.

First, as shown in FIG. 3A, a thin film 71 made of a material having a different factor from the substrate 70 is formed on a desired portion on an insulating substrate 70. Then, when a thin film made of an appropriate material is deposited under appropriate deposition conditions, the deposited thin film 72 grows only on the thin film 71 and does not grow on the substrate 70, as shown in FIG. Can be generated. By utilizing this phenomenon, the deposited thin film 72 formed in a self-aligned manner can be grown, and the conventional lithography process using a resist can be omitted.

As a material that can be deposited by such a selective nucleation crystal growth method, for example, as the substrate 70, Si can be used.
O ₂ , the thin film 71 is Si, GaAs, silicon nitride, and the thin film 72 to be deposited is Si, W, GaAs, InP, or the like.

FIG. 12 is a graph showing changes over time in the nucleation density when Si is deposited on the SiO ₂ and silicon nitride deposition surfaces.

As the graph shows, shortly after the Si deposition started
The nucleation density on SiO ₂ saturates below 10 ³ cm ^{-2 and} its value hardly changes even after 20 minutes. On the other hand, on silicon nitride (Si ₃ N ₄ ), it saturates at about 4 × 10 ⁻⁵ cm ⁻² and then does not change for about 10 minutes, but it increases rapidly thereafter. In addition, in this measurement example, SiCl ₄ gas is changed to H ₂
Dilute with gas and CVD under pressure 175 Torr and temperature 1000 ℃
It shows the case of deposition by the method. Besides, SiH ₄ , SiH ₂ C
Similar effects can be obtained by adjusting the pressure, temperature, etc., using l ₂ , SiHCl ₃ , SiF _4, etc. as the reaction gas. Also, vacuum deposition is possible.

In this case, nucleation on SiO ₂ is hardly a problem, but by adding HCl gas to the reaction gas, nucleation on SiO ₂ is further suppressed, and Si is not deposited on SiO _2. Can be

This phenomenon is largely due to the difference in the adsorption coefficient, desorption coefficient, surface diffusion coefficient, etc. of Si on the material surfaces of SiO ₂ and silicon nitride, but SiO ₂ reacts with the Si atoms themselves and the vapor pressure is high. The fact that SiO ₂ itself is etched by the generation of silicon oxide and such an etching phenomenon does not occur on silicon nitride is also considered to be the cause of selective deposition (T. Yoneha
ra, S.Yoshioka, S.Mayazawa, Journal of Applied Physic
s53,6893,1982).

Thus, by selecting SiO ₂ and silicon nitride as the material of the deposition surface and Si as the deposition material, a sufficiently large difference in nucleation density can be obtained as shown in the graph. Although SiO ₂ is preferable as the material of the deposition surface forming the non-nucleation surface, the nucleation density difference can be obtained not only by SiO ₂ but also by changing the value of X as SiO _x . Of course, the material is not limited to these materials, and the difference in nucleation density is 10 times or more the density of nuclei, preferably
It is only necessary to be 10 ³ times or more, and the deposited film can be sufficiently formed selectively even by using the materials as exemplified above.

As another method for obtaining this difference in nucleation density, Si or N or the like may be locally ion-implanted into SiO ₂ to form a region having excessive Si or N or the like.

In this way, by utilizing the difference in nucleation density, by forming a nucleation surface made of a heterogeneous material having a sufficiently higher nucleation density than that of the material forming the non-nucleation surface to be fine enough to grow only a single nucleus, It is possible to form a single nucleus only in a portion where a nucleation surface having a fine area is present, and selectively grow a single crystal from this nucleus.

Since the selective growth of a single crystal is determined by the electronic state of the surface of the deposition surface, especially the state of dangling bonds, the material with low nucleation density (for example, SiO ₂ ) does not need to be a bulk material. It suffices that it is formed only on the surface of an arbitrary material or substrate to form the above-mentioned deposition surface.

Next, a method for forming a Si single crystal thin film having a grain size according to the size of the device region on the substrate will be described with reference to FIG.

As shown in FIG. 13 (A), a non-nucleation surface 2 made of a material having a low nucleation density is formed on an insulating base substrate 1, and a nucleation surface 3 having a high nucleation density is thinly deposited thereon. Then, by patterning by lithography or the like, the nucleation surface 3 is formed sufficiently finely with a distance l.
The size of the nucleation surface 3 depends on the type of material,
It may be a few microns or less.

On this nucleation surface 3, nuclei of a Si single crystal grow while maintaining a single crystal structure to form island-shaped single crystal grains 4 as shown in FIG. 13 (B). In order to form island-shaped single crystal grains 4,
It is necessary to determine the conditions so that no nucleation occurs on the non-nucleation surface 2.

The island-shaped single crystal grains 4 grow further and come into contact with the adjacent single crystal grains 4 as shown in FIG. 13 (C), but the crystal orientations in the plane of the substrate are not constant. A grain boundary 5 is formed at an intermediate position.

Then, the single crystal grains 4 grow three-dimensionally, but since the crystal planes with a slow growth rate appear as facets, the surface is flattened by etching or polishing, and the grain boundaries 5 are removed. As shown in FIG. 13 (D), a single crystal thin film 6 not containing the grain boundaries 5 is formed in a lattice shape.
The size of the single crystal thin film 6 is determined by the interval 1 of the nucleation surface 3 as described above. That is, the nucleation surface 3
The position of the grain boundary can be controlled by appropriately determining the formation pattern of, and a single crystal having a desired size can be formed in a desired arrangement.

As the base substrate 1, fused quartz, heat resistant glass, alumina, ceramics or the like can be used. Also, a substrate whose surface is coated with SiO ₂ , silicon nitride or the like can be used. Conditions for forming a single nucleus of Si are, for example, SiO ₂ formed by atmospheric pressure CVD method as the non-nucleation surface 2 made of a material having a low nucleation density, and decompressed as the nucleation surface 3 having a high nucleation density. When Si ₃ N ₄ formed by the CVD method is used, the source gas is SiCl ₄ , diluted with H ₂ gas, the gas pressure is 175 Torr, and the substrate temperature is 1000 ° C.

Next, an embodiment of the present invention using such a technique will be described with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a photoelectric conversion device according to the present invention, in which various electronic circuits including an optical sensor are formed on a long insulating substrate 1 by utilizing the above-mentioned selective nucleation crystal growth method. ing.

That is, an optical sensor unit 10 for photoelectrically converting document image information, a storage capacitor unit 11 for storing photoelectrically converted charges, a shift register 12, and charges stored in the capacitor unit 11 controlled by the shift register 12. A transfer switch unit 13, an AD conversion unit 14, a correction data memory 15, a correction table memory 16, a γ correction memory 17, and a clock generator 18 for transmitting the signal are formed on the substrate 1.

FIG. 2 and FIG. 3 show an example of forming a device on the Si single crystal thin film 6 produced by the selective nucleation crystal growth method described above by using a known semiconductor integrated circuit technique.

2 (A) and 2 (B) are a plan view showing a photodiode 20 as a photosensor and a storage capacitor 21 and a sectional view taken along the line BB '.

In the figure, a non-nucleating surface ₂ made of SiO _{2 having} a low nucleation density is formed on an insulating substrate 1 made of glass or ceramics, and a high nucleation density made of Si ₃ N ₄ is formed thereon. A Si single crystal thin film 6 is formed on the nucleation surface 3.
Are formed. A photodiode 20 and a storage capacitor 21 are formed on the thin film 6 by using a known semiconductor integrated circuit technology. 22 and 23 are Si
An insulating layer such as O ₂ or Si ₂ N ₄ , 24 is a wiring such as aluminum, and 25 is a crystal grain boundary.

FIGS. 3A and 3B are a plan view of the CMOS inverter and a cross-sectional view taken along the line BB '.

In the figure, 30 is a p-channel transistor, 31 is n
It is a channel transistor.

In this way, the photoelectric conversion device shown in FIG. 1 is configured by the various devices configured as shown in FIGS. 2 and 3.

Although the photodiodes, transistors, and the like are formed by the normal MOS process in the above-described embodiments, the device configuration is not limited to this, and it goes without saying that all devices that can be normally formed on a silicon wafer, such as a bipolar process, can be formed. Nor. Although the photodiode is used as the light receiving element, a photoconductive material such as amorphous silicon, GdS, CdSe may be used.

FIG. 4 and FIG. 5 are a block diagram of the photoelectric conversion device shown in FIG. 1 and an equivalent circuit diagram of one pixel.

In both figures, the image information of the document is converted into a photocurrent by the photosensor unit 10, and the photocurrent is stored in the storage capacitor unit 11 as a stored charge for a certain period of time. The accumulated charges are output through the transfer switch unit 12 driven by the shift register 13 and input to the AD converter 14.

The correction data memory 15 has two kinds of operations, one is a reference value writing operation for writing the output of the optical sensor unit 10 that has read a reference white original, and the other is each light for actual image reading. It is a reference value reading operation for correcting the variation of the sensor.

Timings of the reference value writing operation and the reading operation are shown in FIGS. 6 and 7, respectively.

First, the reference value writing operation will be described. The optical information of the reference white original or the reference white reading section is photoelectrically converted by the optical sensor section 10 and stored in the storage capacitor section 11 as electric charge. This charge is sequentially sent out as an output signal SG from the transfer switches T _{1 to} T _n which are sequentially turned on by the shift pulse CP1.

The signal SG is input to the AD converter 14, and its conversion output is
Input signals D _{0 to} D _m to the correction data memory 15 via the switch SW.
Supplied as. The shift palace CP2 is generated t time after the conversion by the AD converter 14, and the data at the time of reading the reference white is written in the correction data memory 15.

Next, the reference value reading operation will be described. Pixel information of the document are sequentially output as the signal SG in accordance with a shift pulse CP1, it is converted by the AD converter 14 into a digital signal A ₀ to A _m. On the other hand, the shift pulse CP2 is also generated at the same timing as CP1 and outputs the reference white data Q _{0 to} Q _m stored in the correction data memory 15 at the time of reading the reference white in correspondence with the pixels which are AD-converted.

Signal A ₀ to A _m and the data Q ₀ to Q _m is inputted as the address of the correction table memory 16. This correction table memory 16 is Is a conversion table for performing the calculation. That is, the correction data C _{0 to} C _m are output using the signals A _{0 to} A _m and the data Q _{0 to} Q _m as addresses.

The correction data C _{0 to} C _m thus obtained are stored in the γ correction memory.
Entered as an address in 17. The γ correction memory 17 is used for the purpose of correcting the light amount dependent characteristic that can be obtained by the characteristic of the optical sensor and improving the image quality by adjusting the density characteristic of the document. Several types of conversion patterns are prepared in the γ correction memory 17 and can be switched according to the original.

In the above embodiment, the correction data memory for correcting variations is composed of the RWM, but it may be composed of a UV erasable ROM, or a so-called one-time ROM for writing by destroying fuses, diodes, etc. . In this case, after the substrate process is completed, the characteristics can be measured and permanently written in the correction data memory according to the variation amount of each optical sensor.

[Effects of the Invention] As described in detail above, an optical sensor, a transfer switch, a shift register, a correction memory, and the like are formed on a long substrate by using a technique of selectively growing single crystal silicon on an amorphous substrate. By forming it monolithically, the following effects can be obtained.

(1) Since an amorphous substrate is used, the size of the substrate is not limited and the length can be easily increased.

(2) Since the device can be composed of a crystalline silicon device, the performance is almost the same as that of a device manufactured using a normal crystalline silicon wafer. For this reason, high-speed operation not possible with sensors using polycrystalline or amorphous silicon until now (for example, when used in a facsimile, an A4 size original
Read in ~ 2 seconds).

(3) Since the correction memory is built in the substrate and the reading can be performed by the clock control equivalent to the signal reading from the optical sensor, the clock generation circuit can be simplified.

(4) By using the correction memory as a ROM, the sensitivity variation data of the optical sensor can be permanently stored in the substrate. As a result, the optical sensor and its variation data can be handled on the same substrate, which facilitates management of device assembly and the like.

(5) Since the AD conversion unit and each correction memory unit can be manufactured on the same substrate, a weak analog signal does not go out of the substrate. Therefore, the influence of external noise is small and a high S / N ratio can be realized.

(6) All devices on the substrate can be formed simultaneously by a series of IC manufacturing processes, resulting in high yield and low cost.

[Brief description of drawings]

1 is a schematic configuration diagram showing an embodiment of a photoelectric conversion device according to the present invention, FIGS. 2 and 3 are plan views and cross-sectional views of devices constituting the photoelectric conversion device shown in FIG. 1, FIG. 4 is a block diagram of this embodiment, FIG. 5 is an equivalent circuit diagram of one pixel of this embodiment, and FIGS. 6 and 7 are timing charts for explaining the operation of this embodiment. FIG. 8 is a schematic configuration diagram showing a conventional image reading device, FIG. 9 is a schematic configuration diagram of a photoelectric conversion unit shown in FIG. 8, and FIG. 10 is an equivalent circuit diagram of FIG. FIG. 11 is an explanatory view of the selective nucleation crystal growth method, FIG. 12 is a graph showing the change over time in the nucleation density by the selective nucleation crystal growth method, and FIG. 13 is a selective nucleation crystal growth method. It is explanatory drawing of the method of forming a single crystal Si thin film on an insulating substrate using the method. 1 ... Substrate 2 ... Non-nucleation surface 3 ... Nucleation surface 6 ... Single crystal thin film 10 ... Optical sensor part 11 ... Storage capacitor part 12 ... Shift register 13 ... Transfer switch part 14 ... AD conversion Device 15 …… Correction data memory 16 …… Correction table memory 17 …… γ correction memory

Claims (2)

(57) [Claims] 1. A non-nucleation surface having a low nucleation density and a nucleation surface having an area sufficiently smaller than that of a single nucleus for crystal growth and having a nucleation density higher than that of the non-nucleation surface. An optical sensor unit is provided on a semiconductor single crystal grown from the single nucleus by subjecting a substrate having a free surface arranged adjacently to a crystal formation process, and for outputting signals of the optical sensor unit in time series. The photoelectric conversion device, wherein the transfer switch section, the drive section for driving the transfer switch section, and the correction memory section for storing the variation information of the sensitivity of each optical sensor element of the optical sensor section are formed. 2. The photoelectric conversion device according to claim 1, wherein the correction memory unit is a read-only memory.