JPH0449264B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a photoelectric conversion method, and more particularly to a photoelectric conversion method in which carriers generated by incident light are accumulated and signals are read out based on the accumulated carriers.

[Prior Art] In recent years, research on photoelectric conversion devices, particularly solid-state imaging devices, has been actively conducted along with the progress of semiconductor technology, and some of them have begun to be put into practical use.

These solid-state imaging devices can be broadly classified into
It is classified into two types: CCD type and MOS type. A CCD type imaging device forms a potential well under a MOS capacitor electrode, stores charges generated by incident light in this well, and during readout, these potential wells are sequentially moved by pulses applied to the electrode. The principle is that the accumulated charge is transferred to the output amplifier section and read out. Also
Some CCD imaging devices use a pn junction diode structure for the light receiving section and a CCD structure for the transfer section. On the other hand, in a MOS type imaging device, charges generated by incident light are accumulated in each photodiode made of a pn junction that constitutes the light receiving section, and when reading out, the MOS switching transistor connected to each photodiode is activated. It uses the principle that the accumulated charge is read out to the output amplifier section by sequentially turning on the transistors.

CCD type imaging device has a relatively simple structure,
In addition, considering the noise that can be generated, only the capacitance value of the charge detector consisting of the floating diffusion in the final stage contributes to random noise, so it is a relatively low-noise imaging device and is capable of low-light photography. be. However, due to process constraints in manufacturing CCD type imaging devices, MOS is used as the output amplifier.
Since the type amplifier is on-chip, 1/f noise, which is easily noticeable on images, is generated from the interface between silicon and SiO ₂ film. Therefore, although it is said to have low noise, there are limits to its performance. Furthermore, if the number of cells is increased to achieve higher density in order to achieve higher resolution, the maximum amount of charge that can be stored in one potential well will decrease, making it impossible to maintain a dynamic range. This becomes a big problem as resolution increases. Furthermore, since a CCD-type imaging device transfers accumulated charge by sequentially moving the potential wells, even if there is a defect in one of the cells, charge transfer may stop at that point, or It also has the disadvantage that the manufacturing yield cannot be improved.

On the other hand, MOS type imaging devices are structurally
Although it is a little more complicated than a CCD type imaging device, especially a frame transfer type device, it has the advantage of being able to be configured to have a large storage capacity and having a wide dynamic range. Further, even if a defect exists in one of the cells, because of the XY addressing method, the defect does not affect other cells, which is advantageous in terms of manufacturing yield. However, this
In MOS type imaging devices, wiring capacitance is connected to each photodiode during signal readout, which causes an extremely large signal voltage drop and a drop in the output voltage.The wiring capacitance is large, which causes random noise. It also has disadvantages such as fixed pattern noise mixed in due to variations in the parasitic capacitance of each photodiode and MOS switching transistor for horizontal scanning, and that low-light photography is difficult compared to CCD type imaging devices. have.

Furthermore, in the future, as the resolution of imaging devices increases, the size of each cell will be reduced and the amount of accumulated charge will decrease. On the other hand, the wiring capacitance, which is determined by the chip size, does not decrease much even if the line width is made thinner. For this reason, the MOS type imaging device becomes increasingly disadvantageous in terms of S/N.

Although CCD type and MOS type imaging devices have the above-mentioned advantages and disadvantages, they are gradually approaching the level of practical use. However, it can be said that there are essentially major problems in promoting the higher resolution that will be required in the future.

On the other hand, regarding solid-state imaging devices,
−150878 Publication “Semiconductor Imaging Device”, Japanese Unexamined Patent Publication No. 1983-
Publication No. 157073 “Semiconductor Imaging Device” Japanese Patent Application Laid-open No. 1983-
A new method is proposed in Publication No. 165473 "Semiconductor imaging device". CCD type and MOS type imaging devices are
Charges generated by incident light are transferred to the main electrode (e.g.
In contrast, the method proposed here stores the charge generated by incident light on the control electrode (e.g., the base of a bipolar transistor, the SIT (static induction transistor) or the MOS transistor). gate)
It is based on a new concept of controlling the flowing current using the charges accumulated in the light and generated by light. In other words, whereas the CCD type and MOS type read out the accumulated charge itself to the outside, the method proposed here amplifies the charge using the amplification function of each cell and then reads out the accumulated charge. If you look at it from another perspective, it is read out as a low impedance output through impedance conversion. Therefore, the method proposed here is suitable for high output, wide dynamic range,
Because it has low noise and the carrier (charge) excited by the optical signal accumulates on the control electrode,
It has several advantages such as non-destructive readout. Furthermore, it can be said that this method has the potential for higher resolution in the future.

[Technical Problems to be Solved by the Invention] However, this method is basically an X-Y addressing method, and the element structure described in the above publication is a bipolar transistor in each cell of a conventional MOS type imaging device. The basic configuration is a combination of amplification elements such as , SIT transistors, etc.
Therefore, it has a relatively complicated structure, and although it has the possibility of achieving high resolution, there is a limit to how high resolution can be achieved as it is.

There are also limitations in the points described below. JP-A-56-150788 mentioned above, JP-A-56-
No. 157073, Japanese Patent Application Laid-open No. 165473, and “Application to SIT (Static Injection Transistor) Image Sensor, Technical Report of the Television Society (hereinafter referred to as
``TV Society Journal'' is a typical example of the prior art in which one of the inventors of the present invention was involved.

JP-A-56-150878 and JP-A-56-157073 disclose that charges are added to the P ⁺ region of a hook structure consisting of N ⁺ , P ⁺ , I (or P ^- , N ^- ), and N ⁺ regions. accumulate,
N ⁺ forming a capacitor with ground potential
A configuration is described in which the potential of a region is read out using a switching transistor.

However, since this configuration does not perform reverse bias accumulation, the saturation voltage cannot be arbitrarily set, and since the base potential is not controlled, the output signal cannot be read out at high speed and with sufficient linearity.
Furthermore, in the reset operation after reading, the P ⁺ region is simply grounded without even resetting the output side, resulting in many noticeable afterimages and large fixed pattern noise.

On the other hand, Japanese Patent Application Laid-open No. 56-165473 describes the N ⁺ region,
N + , P ⁺ , I (or P ^{- ,} N ^- ), N ⁺ , consisting of a floating P ⁺ region, a high resistance region and an N ⁺ region connected to a transparent electrode to which a pulsed voltage is applied. The hook structure of the region is shown. The floating N ⁺ region also serves as one of the main electrode regions of the readout transistor, and during readout, the transistor is turned on and electrons flow into the positively charged N ⁺ region, and the voltage change is used as a signal. Perform reading. However, this is also
Similar to the photoelectric conversion devices described in JP-150878 and JP-56-157073, the saturation voltage cannot be set arbitrarily, and the output signal cannot be read out at high speed and with sufficient linearity. Further, in the reset operation after reading, the N ⁺ region on the transparent electrode side is simply set to 0 or a slightly negative potential. In this case, since the base potential is not controlled, high-speed refreshing is impossible, and since there is no reset on the output side, many noticeable afterimages occur. Fixed pattern noise is also large.

The TV Society Journal describes gate storage type photocells and base storage type photocells. Among these, the gate storage type photocell has a configuration in which the gate is placed in a floating state, the gate region is reverse biased to a predetermined voltage via a refresh line via an insulating film, and the voltage is read out to an output circuit with a grounded source resistive load. However, with this configuration, sufficient linearity cannot be obtained when attempting to read out the output signal at high speed. This is because the output voltage does not reach the required value in a short period of time because a sufficient forward bias is not applied during reading. Furthermore, since there is no reset on the output side, the reset operation is insufficient and many afterimages occur.

Moreover, on the other hand, the base storage type photocell has N ⁺ ,
It has a P ⁺ , N ^- , N ⁺ phototransistor structure, and includes a floating base (P ⁺ ), a collector (N ⁺ ) to which a pulsed voltage is applied, a capacitor, and a switching MOSFET. It consists of an emitter (N ⁺ ) to which the output circuit of the emitter follower is connected. However, in this configuration, since reading is performed by applying a voltage to the collector, it is difficult to ensure linearity in high-speed operation, as will be described later with reference to FIGS. 5 and 6.

In addition to the above-mentioned conventional technology, there is also a U.S. patent
In the specification of No. 3624428 and Japanese Patent Publication No. 50-38531, an output circuit with a grounded emitter resistor load is connected to a transistor whose base is provided with an electrode via an insulating layer, and the base is reverse biased to perform storage and storage operation. A configuration is shown in which current is read out in the output circuit of the emitter-grounded resistive load. However, since it is a destructive current readout, linearity and afterimage characteristics are poor. Since the base and collector are connected to each other with a relatively high impurity concentration, the sensitivity, especially the blue sensitivity, is low, and the voltage change is read out as a signal. However, this is also
Similar to the photoelectric conversion devices described in Publication No. 150878 and Publication No. 56-157073, the saturation voltage cannot be set arbitrarily, and the output signal cannot be read out at high speed and with sufficient linearity. Further, in the reset operation after reading, the N ⁺ region on the transparent electrode side is simply set to 0 or a slightly negative potential. In this case, since the base potential is not controlled, high-speed refreshing is impossible, and since there is no reset on the output side, many noticeable afterimages occur. Fixed pattern noise is also large.

The TV Society Journal describes gate storage type photocells and base storage type photocells. Among these, the gate storage type photocell has a configuration in which the gate is placed in a floating state, the gate region is reverse biased to a predetermined voltage via a refresh line via an insulating film, and the voltage is read out to an output circuit with a grounded source resistive load. However, with this configuration, sufficient linearity cannot be obtained when attempting to read out the output signal at high speed. This is because the output voltage does not reach the required value in a short period of time because a sufficient forward bias is not applied during reading. Furthermore, since there is no reset on the output side, the reset operation is insufficient and many afterimages occur.

Moreover, on the other hand, the base storage type photocell has N ⁺ ,
It has a P ⁺ , N ^- , N ⁺ phototransistor structure, and includes a floating base (P ⁺ ), a collector (N ⁺ ) to which a pulsed voltage is applied, a capacitor, and a switching MOSFET. It consists of an emitter (N ⁺ ) to which the output circuit of the emitter follower is connected. However, in this configuration, since reading is performed by applying a voltage to the collector, it is difficult to ensure linearity in high-speed operation, as will be described later with reference to FIGS. 5 and 6.

In addition to the above-mentioned conventional technology, there is also a U.S. patent
In the specification of No. 3624428 and Japanese Patent Publication No. 50-38531, an output circuit with a grounded emitter resistor load is connected to a transistor whose base is provided with an electrode via an insulating layer, and the base is reverse biased to perform storage and storage operation. A configuration is shown in which current is read out in the output circuit of the emitter-grounded resistive load. However, since it is a destructive current readout, linearity and afterimage characteristics are poor. Since the base and collector are connected to each other with a relatively high impurity concentration, the sensitivity, particularly the blue sensitivity, is low.

[Object of the Invention] An object of the present invention is to provide a new photoelectric conversion method that has an amplification function in each cell but has an extremely simple structure and can sufficiently cope with future increases in resolution.

Another object of the present invention is to provide a photoelectric conversion method excellent in high speed and capable of obtaining an output signal with good linearity for irradiated light in a very short time.

Still another object of the present invention is to provide an excellent photoelectric conversion method in which afterimages and fixed pattern noise are hardly a problem even when performing high-speed readout with good linearity.

Still another object of the present invention is to provide a photoelectric conversion method that is sensitive even to blue light.

This purpose includes: a first semiconductor region of a first conductivity type that is electrically connected to a first electrode; and a semiconductor region that is disposed adjacent to the first semiconductor region and has a higher resistivity than the first semiconductor region. a second semiconductor region, and a third semiconductor region of a second conductivity type different from the first conductivity type, which is disposed adjacent to the second semiconductor region and is disposed with respect to the second electrode via an insulating region. a semiconductor region; and a fourth semiconductor region of a first conductivity type, which is disposed adjacent to the third semiconductor region and is electrically connected to a third electrode connected to an output circuit including a capacitive load. A photoelectric conversion method using a photoelectric conversion device including a transistor, wherein the fourth semiconductor region is held at a ground potential or in a floating state, and a junction between the fourth semiconductor region and the third semiconductor region is moved in the opposite direction. an accumulation step of biasing and irradiating the transistor with light in order to accumulate carriers generated by receiving light energy in the third semiconductor region; and keeping the fourth semiconductor region in a floating state and and the fourth semiconductor region; a readout step comprising forward biasing a junction between the fourth semiconductor region and the fourth semiconductor region and reading an output signal corresponding to the accumulated carrier from the third electrode as a voltage at the capacitive load; a refresh step that includes holding at ground potential and biasing the junction between the third semiconductor region and the fourth semiconductor region in the forward direction to remove accumulated carriers; and a photoelectric conversion characterized by sequentially performing the following steps This is accomplished by a method.

[Operation] According to the present invention, during each of the storage, readout, and refresh operations, the potential of the third semiconductor region, which serves as the control electrode region, is controlled independently of the fourth semiconductor region, which serves as the main electrode region, and the first semiconductor region. can be controlled. During accumulation, the junction between the third semiconductor region and the fourth semiconductor region connected to the output circuit changes from a reversely biased state to the potential of the third semiconductor region as the charge accumulates to the potential of the fourth semiconductor region. Change. Therefore, since the saturation voltage is determined by the bias voltage at this time, a large saturation voltage can be ensured. At the next read,
The junction can be deeply biased in the forward direction. In this way, it is possible to obtain an output signal with good linearity for the irradiated light in an extremely short period of time. In this case, since the charges accumulated in the third semiconductor region are non-destructive, non-destructive reading is possible, and the cell itself has a switching action, so a reading switching transistor can be omitted. Also, during refresh, the third
By independently controlling the potential of the semiconductor region and deeply biasing the junction between the third and fourth semiconductor regions in the forward direction, refresh can be performed through the fourth semiconductor region connected to the output circuit, resulting in high-speed operation. Afterimages and noise can be removed.

[Example] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a diagram illustrating the basic structure and operation of a photosensor cell that constitutes a photoelectric conversion device according to an embodiment of the present invention.

FIG. 1a shows a plan view of the optical sensor cell, FIG. 1b shows a sectional view of a section AA' in the plan view of FIG. 1a, and FIG. 1c shows an equivalent circuit thereof. In addition, the same numbers are given to the parts common to FIGS. 1a, b, and c in each part.

Although FIG. 1 shows a plan view of the aligned layout method, it goes without saying that the pixel shifting method (interpolation layout method) can also be used to increase the horizontal resolution.

As shown in FIGS. 1a and 1b, this optical sensor cell consists of silicon constituting a collector as a first semiconductor region doped with impurities such as phosphorus P, antimony Sb, and arsenic As to make it n-type or n ⁺ type. On the substrate 1, there are a passivation film 2 usually made of a PSG film or the like; an insulating oxide film 3 made of a silicon oxide film SiO ₂ ; a SiO ₂ or Si ₃ film for electrically insulating between neighboring optical sensor cells. Element isolation region 4 made of an insulating film made of N ₄ or the like or a polysilicon film; n ⁻ as a second semiconductor region formed by epitaxial technology or the like with low impurity concentration and higher resistivity than the substrate 1; Region 5; P region 6, which serves as the base of the bipolar transistor as a third semiconductor region doped with an impurity such as boron B using an impurity diffusion technique or an ion implantation technique; An n ⁺ region 7 which serves as the emitter of a bipolar transistor as a fourth semiconductor region to be formed; is formed of a conductive material such as aluminum Al, Al-Si, Al-Cu-Si, etc. for reading out signals to the outside. Wiring 8 as a third electrode; electrode 9 as a second electrode for applying a pulse to the floating p region 6 through the insulating film 3; wiring 10 therefor; ohmic contact on the back surface of the substrate 1; n ⁺ region 11 with a high impurity concentration formed by impurity diffusion technology etc. to provide a substrate potential, that is, a collector potential of a bipolar transistor as a first electrode formed of a conductive material such as aluminum. Electrode 12: Consisting of:

Note that 19 in FIG. 1A is a contact portion for connecting the n ⁺ region 7 and the wiring 8. Further, the alternating portions of the wirings 8 and 10 are so-called two-layer wirings, and are insulated from each other by insulating regions formed of an insulating material such as SiO ₂ . That is, it has a two-layer metal wiring structure.

The capacitor Cox 13 in the equivalent circuit of ^FIG .
It is composed of an n ^- region 5 having a low impurity concentration and an n or n ⁺ region 1 serving as a collector.
As is clear from these drawings, p region 6 is made into a floating region.

The second equivalent circuit in FIG. 1c shows the bipolar transistor 14 with base-emitter junction capacitance.
Cbe15, base-emitter pn junction diode
It is expressed using Dbe16, base-collector junction capacitance Cbc17, and base-collector pn junction diode Dbc18.

Here, in the equivalent circuit diagram, symbols indicating two current sources in different directions, which should originally be drawn in parallel with the pn junction diode Dbe16 and the pn junction diode Dbc18, are omitted.

Here, the carrier accumulation step, readout step, and refresh step in the photoelectric conversion method of the present invention will be briefly explained.

In the carrier accumulation step, for example, the emitter of a bipolar transistor is grounded, the collector is set to a positive potential, and the base is set to a negative potential to create a reverse bias state with respect to the emitter. When light is irradiated in this state, electron-hole pairs are generated and holes are accumulated in the base region, and the base potential is the potential at which the holes accumulated in the base begin to flow out to the emitter (here, the ground potential). ). In this case, the saturation voltage of the photosensor cell is given by the potential difference between the bias potential when the base is biased to a negative potential and the ground potential, so a large saturation voltage can be obtained by setting the bias potential.

Next, in the readout process, the emitter is in a floating state, the collector is set to a predetermined positive potential, a positive voltage for readout is applied to the electrode 9 in FIG. Signal reading is performed in the bias state. Note that the holes accumulated in the base have an extremely low recombination probability in the p region 6 of FIG. 1b, so non-destructive readout is possible.

Next, in the refresh step, the emitter is grounded, the collector is grounded or at a positive potential, a positive voltage is applied to the electrode 9 in FIG. 1b, the base potential is increased, and the emitter is forward biased. Lower the base potential. This operation refreshes the charge accumulated in the base.
Note that when the positive voltage applied to the electrode 9 in FIG. 1b is returned to zero volts, the base potential becomes a negative potential, resulting in the reverse bias state described in the charge accumulation step.

The circuit configuration will be explained with reference to FIGS. 2 and 4. An output circuit is connected to the emitter region, which is the fourth semiconductor region, of a photoelectric conversion cell including a transistor as shown by the reference numeral 30 in FIG. There is. This output circuit consists of vertical lines 38, 38', 3
8", horizontal shift register 39, MOS transistors 40, 40', 40", output line 41, MOS
It consists of a transistor 42, an output transistor 44, a load resistor 45, etc., and vertical lines 38, 38',
38'' each has a wiring capacitance as a capacitive load.

Further, a vertical shift register 32, a buffer MOS transistor 33, a vertical shift register 32, a buffer MOS transistor 33,
33', 33'', terminal 34, horizontal line 31, 3
1', 31'' is provided.

The potential of the base region (third semiconductor region), which is a control electrode region, is controlled by a readout means independently of the emitter and collector regions (fourth semiconductor region, first semiconductor region), which are main electrode regions. Ru. The operation at this time will be explained with reference to FIG. At the time of reading, a positive voltage V _R is independently applied from the wiring 10 to the emitter as a main electrode region in a floating state and the collector as a main electrode region held at a positive potential, thereby changing the emitter potential. On the other hand, by increasing the base potential, the forward potential is positively biased for the junction between the control electrode region and the main electrode region connected to the output circuit. In this way, a current flows until the emitter potential becomes equal to the base potential, that is, the accumulated voltage generated by light irradiation, but the time required at this time is shortened by the action of the voltage _VR , even in high-speed readout. Excellent linearity can be ensured.

On the other hand, the emitter as the first main electrode region is connected to the MOS transistors 48 and 4 as switching means.
8', 48'' is connected to a first reference voltage source indicated with a ground symbol and grounded.

At this time, the collector as the second main electrode region is connected to the second reference voltage source, that is, set to a positive potential or ground potential.In this state, a voltage of positive potential _VRH is applied to the collector as the control electrode region. By controlling the potential, at least the base-emitter region is forward biased, holes accumulated in the base region flow out, electrons flow into the base region, and the accumulated charges disappear. As a forward bias means for applying such a forward bias, in addition to the above reading means, for example, the buffer MOS transistors 35 and 3 shown in FIG.
5', 35'', terminals 36, 37, etc.

The basic operation of the optical sensor cell will be explained below with reference to FIG.

The basic operation of this photosensor cell consists of a charge accumulation operation by light incidence, a readout operation, and a refresh operation.

First, charge storage operation will be explained.

In charge storage operation, for example, the emitter is
It is grounded through a wiring 8, and its collector is biased to a positive potential through a wiring 12. It is also assumed that the base is brought to a negative potential, that is, to a reverse bias state with respect to the emitter 7, by applying a positive pulse voltage to the capacitor Cox13 through the wiring 10 in advance. The operation of biasing the base 6 to a negative potential by applying a pulse to Cox 13 will be explained in detail later when the refresh operation is explained.

In this state, when light 20 enters from the front side of the photosensor cell as shown in FIG. 1, electron-hole pairs are generated within the semiconductor.
Of these, electrons flow toward the n-region 1 side because the n-region 1 is biased to a positive potential, but holes are rapidly accumulated in the p-region 6. Due to the accumulation of holes in the p region, the potential of the p region 6 gradually changes toward a positive potential.

In Figures 1a and 1b, the lower surface of the light-receiving surface of each sensor cell is mostly occupied by the p region, and some n ⁺
It has become area 7. Naturally, the concentration of electron-hole pairs excited by light increases as it approaches the surface. Therefore, many electron-hole pairs are excited in the p-region 6 by the light.
If the structure is such that electrons photoexcited in the p-region immediately flow out of the p-region 6 without recombination and are absorbed in the n-region, the holes excited in the p-region 6 will be accumulated as they are. p
The region 6 is changed to a positive potential direction. When the impurity concentration of p region 6 is uniform, electrons excited by light are diffused and are connected to p region 6 and n ^-
It flows to the pn ^- junction with region 5, and is then absorbed by n collector region 1 due to drift due to the strong electric field applied to the n ^- region. Of course, it is possible for electrons to travel within the p region 6 by diffusion alone, but if the structure is configured so that the p-based impurity concentration decreases from the surface to the inside, this impurity concentration difference , an electric field Ed, Ed=1/W _B・kT/q・lnN _AS /N _Ai , is generated in the base from the inside toward the surface. Here, W _B is the depth from the light incident surface of the p region 6, k is the Boltzmann constant, T is the absolute temperature, q is the unit charge, N _AS is the surface impurity concentration of the p base region 6, and N _Ai is the p Region 6 n ^- High resistance region 5
is the impurity concentration at the interface with

Here, if N _AS /N _Ai >3, the movement of electrons in p region 6 will be performed by drift rather than diffusion. That is, in order to effectively operate carriers excited by light in p-region 6 as signals, the impurity concentration of p-region 6 must decrease from the light-incidence side surface toward the inside. is desirable. When p region 6 is formed by diffusion, its impurity concentration decreases toward the inside compared to the light incident side surface.

A portion below the light-receiving surface of the sensor cell is occupied by the n ⁺ region 7 . The depth of n ⁺ region 7 is typically 0.2
Since the thickness is designed to be about 0.3 μm or less, the amount of light absorbed by the n ⁺ region 7 is not so large to begin with, so there is no problem. just,
For light on the short wavelength side, especially blue light, the presence of the n ⁺ region 7 causes a decrease in sensitivity. The impurity concentration of the n ⁺ region 7 is usually designed to be about 1×10 ²⁰ cm ⁻³ or more. The diffusion distance of holes in the n ⁺ region 7 doped with impurities at such a high concentration is about 0.15 to 0.2 μm. Therefore, in order for the holes optically excited in the n ⁺ region 7 to effectively flow into the p region 6, the n ⁺ region 7 must also have a structure in which the impurity concentration decreases from the light incident side surface toward the inside. is desirable. If the impurity concentration distribution of the n ⁺ region 7 is as described above, a strong drift electric field will be generated from the light incident side surface to the inside, and the n ⁺ region 7 will be
The holes that are photoexcited in this direction simply flow into the dust p region 6 due to the drift. If the impurity concentration of both the n ⁺ region 7 and the p region 6 decreases from the light incident side surface toward the inside, then the n ⁺ region 7 and the p region existing on the light incident side surface side of the sensor cell All of the optically excited carriers in 6 function effectively as optical signals. If this n ⁺ region 7 is formed by impurity diffusion from a silicon oxide film or polysilicon film doped with As or P at a high concentration, it is possible to obtain an n ⁺ region with the desired impurity gradient as described above. It is.

Eventually, the accumulation of holes will cause the base potential to change to the emitter potential, in this case to ground potential, where it will be clipped.
More precisely, the base and emitter are biased deeply in the forward direction, and the holes accumulated in the base are clipped at a voltage that begins to flow to the emitter. That is, the saturation potential of the photosensor cell in this case is approximately given by the potential difference between the bias potential when p region 6 is initially biased to a negative potential and the ground potential. When n ⁺ region 7 is not grounded and accumulates charges generated by optical input in a floating state, p region 6 can accumulate charges to approximately the same potential as n region 1 .

The above is a qualitative and general explanation of the charge accumulation operation, but a more specific and quantitative explanation will be given below.

The spectral sensitivity distribution of this optical sensor cell is given by the following equation.

S(λ)=λ/1.24・exp(−αx) x{1−exp(−αy)}・T [A/W] However, λ is the wavelength of light [μm], and α is the light wavelength in the silicon crystal. Attenuation coefficient [μm ^-1 ], x is the thickness of the “dead 1ayer” (dead layer) on the semiconductor surface that causes recombination loss and does not contribute to sensitivity [μm], y is the thickness of the epitaxial layer [μm] , T represent the transmittance, that is, the ratio of the amount of light that effectively enters the semiconductor, taking into account reflection and the like with respect to the amount of incident light. The spectral sensitivity S of this optical sensor cell
(λ) and irradiance Ee (λ) to determine the photocurrent Ip
is calculated using the following formula.

Ip=∫ ^∞ _p S(λ)・Ee(λ)・dλ [μA/cm ² ] However, the irradiance Ee(λ) [μW・cm ⁻²・nm ⁻¹ ] is given by the following formula.

Ee(λ)=E _V・P(λ)/6.80∫ ^∞ / _p V(λ)P(λ
)・dλ [μW・cm ^-2・nm ^-1 ] However, E _V is the illuminance of the sensor's light receiving surface [Lux], P
(λ) is the spectral distribution of light incident on the light receiving surface of the sensor, and V(λ) is the relative luminous efficiency of the human eye.

Using these formulas, for an optical sensor cell with an epitaxial layer of 4 μm, when irradiated with A light source (2854〓) and the sensor light receiving surface illuminance is 1 [Lux], approximately
A photocurrent of 280 nA/cm ^-2 flows, and the number of incident photons or the number of generated electron-hole pairs is about 1.8 x 10 ¹² /cm ² Sec.

Also, at this time, the potential Vp generated by the accumulation of holes excited by light in the base is Vp=
It is given by Q/C. Q is the amount of accumulated hole charge, and C is the junction capacitance that is the sum of Cbe15 and Cbc17.

Now, the impurity concentration of n ⁺ region 7 is 10 ²⁰ cm ^-3 , the impurity concentration of p region 6 is 5 × 10 ¹⁶ cm ^-3 , the impurity concentration of n ^- region 5 is 10 13 cm -3 , and the impurity concentration of n ⁺ region 7 is 10 ¹³ cm ^-3 . The area is 16μm ² , p
When the area of region 6 is 64 μm ² and the thickness of n ^- region 5 is 3 μm, the junction capacitance is about 0.014 pF. On the other hand, the number of holes accumulated in p region 6 is as follows: 60sec, the effective light-receiving area, that is, the product of the area of p-region 6 minus the product of electrodes 8 and 9, is
If it is about 56μm ² , it will be 1.7×10 ⁴ pieces. Therefore, the potential Vp generated by light incidence is about 190 mV.

What should be noted here is that as the resolution increases and the cell size decreases, the amount of light incident on each photosensor cell decreases, and the amount of accumulated charge Q also decreases. Since the junction capacitance decreases in proportion to the cell size as the cell size decreases, the potential Vp generated by light incidence is kept almost constant. This is because the optical sensor cell according to the present invention has an extremely simple structure, as shown in FIG. 1, and has the possibility of having an extremely large effective light-receiving surface.

This is one of the reasons why the photoelectric conversion device of the present invention is advantageous compared to the case of an interline type CCD. The area of the transfer section to be secured becomes relatively large, and as a result, the effective light-receiving surface decreases, resulting in a decrease in sensitivity, that is, the voltage generated by light incidence. In addition, in an interline type CCD image pickup device, the saturation voltage is limited by the size of the transfer section and gradually decreases over time, whereas in the optical sensor cell of the present invention, as mentioned earlier, the saturation voltage is limited by the size of the transfer section and gradually decreases. The saturation voltage is determined by the bias voltage when p-region 6 is biased to a negative potential, and a large saturation voltage can be ensured.

The operation of reading out the voltage generated by the charges accumulated in p region 6 as described above will now be described.

In the read operation state, the emitter and wiring 8 are kept in a floating state, and the collector is held at a positive potential Vcc.

Figure 2 shows an equivalent circuit.

Also here, symbols indicating two current sources in different directions that should originally be written in parallel with the pn junction diode Dbe16 and the pn junction diode Dbc18 as an equivalent circuit are omitted.

Now, if the potential when the base 6 is biased to a negative potential before light irradiation is -V _B , and the accumulated voltage generated by light irradiation is V _P , then the base potential is
The potential is −V _B +V _P. In this state, a positive voltage V _R for reading is applied to the electrode 9 through the wiring 10.
is applied, this positive potential V _R increases the oxide film capacitance
The capacitance is divided by Cox13, the base-emitter junction capacitance Cbe15, and the base-collector junction capacitance Cbc7, and the voltage Cox/Cox+Cbe+Cbc·V _R is added to the base. Therefore, the base potential becomes −V _B +V _P +Cox/Cox+Cbe+Cbc·V _R. If the condition -V _B +Cox/Cox+Cbe+Cbc·V _R =0 is established here, the base potential becomes exactly the accumulated voltage V _P generated by light irradiation. When the base potential is biased in a positive direction with respect to the emitter potential in this way, electrons are injected from the emitter to the base, and since the collector potential is positive, they are accelerated by the drift electric field and flow into the collector. reach. The current flowing at this time is given by the following equation.

i=A _j・q・Dn・n _pe /W _B (1+lnN _Ae /N _AC ) × {expq/kT (V _P −V _e )−1} However, A _j is the junction area between the base and emitter, q
is the unit charge (1.6×10 ^-19 coulombs), D _o is the electron diffusion constant in the base, n _pe is p
Electron concentration as a minority carrier at the base emitter end, W _B is the base width, N _Ae is the acceptor concentration at the base emitter end, N _Ac
is the acceptor concentration at the collector end of the base,
k is Boltzmann's constant, T is absolute temperature, and V _e is emitter potential.

This current has an emitter potential V _e as a base potential,
In other words, here the accumulated voltage generated by light irradiation
It is clear from the above equation that the flow continues until it becomes equal to V _p . At this time, the temporal change in the emitter potential V _e is calculated using the following formula.

Cs・dV _e /dt= i=A _j・q・Dn・n _pe /W _B (1+lnN _Ae /N _Ac × {expq/kT (V _p −V _e )−1} However, here, the wiring capacitance Cs is This is the capacitance 21 of the wiring 8 connected to the emitter.

FIG. 5 shows an example of a temporal change in the emitter potential calculated using the above formula.

According to FIG. 5, it takes about 1 second for the emitter potential to become equal to the base potential. This is because when the emitter potential V _e approaches V _p , less current flows. Therefore, the means to solve this problem is to set the condition -V _B +Cox/Cox+Cbe+Cbc・V _R =0 when applying the positive voltage V _R to the electrode 9, but instead of this condition, A possible method is to insert the condition -V _B +Cox/Cox+Cbe+Cbc・V _R =V _Bias , and bias the base potential by an extra amount of V _Bias in the forward direction.
The current flowing at this time is given by the following equation.

i=A _j・q・Dn・n _pe /W _B (1+lnN _Ae /N _Ac × {expq/KT (V _p +V _Bias −V _e )−1} In Figure 6a, V _Bias = 0.6V In this case, after a certain period of time, the V _R applied to _the electrode 9 is returned to zero volts, and the flowing current is stopped. In Figure 6a, the readout voltage is plotted by subtracting the gain of the constant potential that depends on the readout time due to the bias voltage component. When the voltage V _R is returned to zero volts, the voltage -Cox/Cox+Cbe+Cbc・V _R is added to the base potential, which is the opposite of when it was applied, so the base potential is the state before applying the positive voltage V _R , i.e. , -V _B , and the emitter is reverse biased, so the current flow stops.According to Figure 6a, the readout time is about 100 ns or more (that is, V _R is applied to electrode 9). If you take the time,
The linearity of the storage voltage V _p and the read voltage is ensured over a range of about four digits, indicating that high-speed read is possible. In Figure 6a, the 45° line is the result when sufficient time is taken for reading,
In the above calculation example, the capacitance Cs of the wiring 8 is set to 4 pF, which is about 300 times larger than the junction capacitance of Cbe + Cbc, which is 0.014 pF. However, the accumulated voltage V _p generated in the p region 6 FIG. 6a shows that the signal is not subjected to any attenuation and is read out very quickly due to the effect of the bias voltage. This is because the amplification function, that is, the charge amplification function, of the photosensor cell according to the above configuration is working effectively.

On the other hand, in conventional MOS type imaging devices, the accumulated voltage V _p is Cj・V _p /(Cj + Cs) (however, Cj is
p-n junction capacitance of the light receiving part of the MOS type image pickup device),
It has a drawback that the read voltage value decreases by about 2 digits. For this reason, in MOS type imaging devices,
Fixed pattern noise due to variations in parasitic capacitance of switching MOS transistors for external readout, or random noise caused by large wiring capacitance, that is, output capacitance, is large, and S/
Although there was a problem that the N ratio could not be maintained, in the photosensor cells having the configurations shown in FIG. Since it is quite large, fixed pattern noise and random noise caused by the output capacitance are relatively small, making it possible to obtain a signal with an extremely good S/N ratio.

Previously, we showed that when the bias voltage V _Bias is set to 0.6 V, linearity of about 4 orders of magnitude can be obtained with a high-speed readout time of about 100 nsec, but the relationship between this linearity, readout time, and bias voltage V _Bias is The calculated results are shown in more detail in Figure 6b.

In Figure 6b, the horizontal axis is the bias voltage V _Bias
, and the vertical axis indicates the readout time.
In addition, the parameters are, when the accumulated voltage is 1mV,
It shows the time dependence until the read voltage reaches 80%, 90%, 95%, and 98% of 1 mV. As shown in Figure 6a, when the storage voltage is 80%, 90%, 95%, and 98% at a storage voltage of 1 mV, it is clear that the values are even better at higher storage voltages. it is obvious.

According to this FIG. 6b, the bias voltage V _Bias is
At 0.6V, the readout time is 0.12μs for the readout voltage to be 80% of the storage voltage, and the readout time to be 90% is
0.27μs, 95% is 0.54μs, 98% is
It can be seen that the time is 1.4μs. Also, the bias voltage
This shows that if V _Bias is made larger than 0.6V, even faster reading is possible. Like this,
Once the readout time and required linearity are determined from the overall design of the imager, the required bias voltage V _Bias can be determined by using the graph of FIG. 6b.

Another advantage of the optical sensor cell having the above configuration is that the holes accumulated in the p region 6 can be read out nondestructively because the probability of recombination of electrons and holes in the p region 6 is extremely small. That is, when the voltage V _R applied to the electrode 9 during readout is returned to zero volts, the potential of the p region 6 becomes the reverse bias state before applying the voltage V _R , and the accumulated voltage V _p generated by light irradiation becomes , it will be preserved as is unless it is exposed to new light. This means that when the optical sensor cell according to the above configuration is configured as a photoelectric conversion device, new functions can be provided in terms of system operation.

The time that the accumulated voltage V _p can be held in the p region 6 is extremely long, and the maximum holding time is rather
It is limited by the thermally generated dark current in the junction depletion layer. In other words, this thermally generated dark current saturates the optical sensor cell. However, in the optical sensor cell according to the above configuration, the region where the depletion layer spreads is the n ^- region 5 which is a low impurity concentration region, and this n ^-
Region 5 has an extremely low impurity concentration of about 10 ¹² cm ^-3 to 10 ¹⁴ cm ^-3 , so its crystallinity is good.
Compared to MOS and CCD type imaging devices, fewer electron-hole pairs are thermally generated. Therefore, the dark current is small compared to other conventional devices. That is, the optical sensor cell according to the above configuration essentially has a structure with low dark current noise.

Next, the operation of refreshing the charges accumulated in p region 6 will be explained.

In the optical sensor cell having the above configuration, as already mentioned, the charges accumulated in the p region 6 are not eliminated by the read operation. Therefore, in order to input new optical information, a refresh operation is required to eliminate the previously accumulated charges. At the same time, it is necessary to charge the potential of p region 6, which is in a floating state, to a predetermined negative voltage.

In the optical sensor cell having the above configuration, the refresh operation is also performed by applying a positive voltage to the electrode 9 through the wiring 10, similarly to the read operation. At this time, the emitter is grounded through the wiring 8. The collector is grounded or at a positive potential through the electrode 12. FIG. 3 shows an equivalent circuit for refresh operation. However, an example is shown in which the collector side is grounded.

When a positive voltage _VRH is applied to the electrode 9 in this state, the oxide film capacitance Cox13,
By capacitance division of base-emitter junction capacitance Cbe15 and base-collector junction capacitance Cbc17, a voltage of Cox/Cox + Cbe + Cbc·V _RH is instantaneously applied as in the previous read operation. This voltage causes the base-emitter junction diode Dbe16 and the base-collector junction diode Dbc18 to be forward-biased and become conductive, current begins to flow, and the base potential gradually decreases.

At this time, the change in the potential V of the base in a floating state is approximately expressed by the following equation.

(Cbe+Cbc)dV/dt=-(i ₁ +i ₂ ) However, i ₁ = Ab(qD _{p poe} /Lp+qD _o n _pe /W _B ) x{exp(q/kTV)-1} i ₂ =AeqD _o n _pe /W _B x {exp(q/kTV)-1} i ₁ is the current flowing through the diode Dbc, and i ₂ is the current flowing through the diode Dbe. A _b is the base area,
Ae is the emitter area, Dp is the hole diffusion constant in the collector, P _oe is the hole concentration in the collector at thermal equilibrium, Lp is the mean free path of holes in the collector, n _pe is the electron concentration in the base at thermal equilibrium It is. i ₂ ,
The current due to hole injection from the base side to the emitter can be ignored because the impurity concentration of the emitter is sufficiently higher than that of the base.

The above equation is an approximation of a stepwise junction, and the actual device deviates from a stepwise junction, and the base is thin and has a complicated concentration distribution, so it is not exact. , the refresh operation can be explained with a fair approximation.

Of the current i ₁ flowing between the base and the collector in the above equation, q·Dp· _poe /Lp represents a current due to holes, that is, a component in which holes flow from the base to the collector side. In order to facilitate the flow of current due to these holes, in the optical sensor cell having the above configuration, the impurity concentration of the collector is designed to be a little lower than that of a normal bipolar transistor.

It was calculated using this formula. FIG. 7 shows an example of the time dependence of the base potential. The horizontal axis shows the passage of time from the moment the refresh voltage V _RH was applied to the electrode 9, that is, the refresh time, and the vertical axis shows the base potential. In addition, the initial potential of the base is used as a parameter. The initial potential of the base is the potential exhibited by the base in a floating state at the moment the refresh voltage V _RH is applied, and V _RH ,
It depends on Cox, Cbe, Cbc and the charges stored in the base.

Looking at FIG. 7, the potential of the base always falls along a straight line on the semi-logarithmic graph after a certain period of time, regardless of the initial potential.

FIG. 7b shows experimental values of base potential change with respect to refresh time. Compared to the calculation example shown in Figure 7a, the test device used in this experiment has a much larger dimension, so although the absolute value does not match the calculation example, the base potential change with respect to the refresh time is semi-logarithmic. It has been demonstrated that it changes linearly on the graph. This experimental example shows the value when both the collector and emitter are grounded.

Now, the maximum value of the accumulated voltage V _p due to light irradiation is 0.4 [V],
If the voltage V applied to the base by the refresh voltage V _RH is 0.4 [V], the maximum value of the initial base potential is 0.8 [V] as shown in Figure 7, and 10 ^-15 [sec] after the application of the refresh voltage. The base potential begins to fall in a straight line, and after 10 ^-5 [sec], it matches the potential change when no light is applied, that is, when the initial base potential is 0.4 [V].

There are two ways in which the p region 6 can be charged to a negative potential by applying a positive voltage for a certain period of time through the MOS capacitor Cox and removing the positive voltage. One is an operation in which holes with positive charges flow from p region 6 to n region 1 which is mainly in a grounded state, thereby accumulating negative charges.
In order to prevent holes from flowing unilaterally from p-region 6 to n-region 1 and to prevent electrons from n-region 1 from flowing into p-region 6 too much, the impurity density of p-region 6 must be set to the impurity density of n-region 1. It should be higher. On the other hand, the electrons from n ⁺ region 1 are
By flowing into the p region 6 and recombining with holes, negative charges can also be accumulated in the p region 6. In this case, the impurity density of n region 1 is higher than that of p region 6. The operation of accumulating negative charges due to holes flowing out from p-region 6 is much faster than the operation of accumulating negative charges due to electrons flowing into the base of p-region 6 and recombining with holes. However, according to experiments conducted so far, it has been confirmed that even the refresh operation in which electrons flow into the p region 6 exhibits a sufficiently fast time response for the operation of the photoelectric conversion device.

When a photoelectric conversion device is constructed by arranging a large number of optical sensor cells according to the above configuration in the XY direction, the image shows that the accumulated voltage V _p of each sensor cell varies between 0 and 0.4 [V] in the above example. , 10 ^-5 [sec] after applying the refresh voltage V _RH , a constant voltage of approximately 0.3 [V] remains at the base of all sensor cells, but all changes in the product voltage V _p due to the image disappear. I understand that. That is, in the photoelectric conversion device using optical sensor cells according to the above configuration, there is a complete refresh mode in which the base potential of all sensor cells is brought to zero volts by a refresh operation (in this case, in the example shown in FIG. 7a),
10 [sec]), and a transient refresh mode in which a certain constant voltage remains at the base potential but the fluctuation component due to the accumulated voltage V _p disappears (in this case, as shown in Figure 7a) In this example, the refresh pulse is 10 [μsec] to 10 [sec]. In the above example, the voltage V _A applied to the base by the refresh voltage V _RH was set to 0.4 [V], but if this voltage V _A is set to 0.6 [V], the above transient refresh mode According to Figure 7, 1
It occurs in nsec and can be refreshed extremely quickly. The choice of whether to operate in the complete refresh mode or the transient refresh mode is determined by the purpose of use of the photoelectric conversion device.

If the voltage remaining at the base in this transient refresh mode is V _R , then the refresh voltage V _RH
After application, in the transient state at the moment when V _RH is returned to zero volts, a negative voltage of -Cox/Cox+Cbe+Cbc・V _RH is added to the base, so the base potential after the refresh operation by the refresh pulse is V _R -Cox/ Cox + Cbe + Cbc・V _RH , and the base becomes reverse biased with respect to the emitter.

It was explained earlier that during the accumulation operation of accumulating carriers excited by light, the base is in a reverse bias state in the accumulation state, but this refresh operation brings the refresh and base to a reverse bias state. These two operations are performed simultaneously.

FIG. 7c shows experimental values of changes in the base potential V _R −Cox/Cox+Cbe+Cbc·V _RH after the refresh operation with respect to the refresh voltage V _RH . The parameter Cox value is set from 5pF to 100pF. The circles are experimental values, and the solid lines are calculated values calculated from V _k −Cox/Cox+Cbe+Cbc·V _RH . At this time
V _k =0.52V, and Cbc+Cbe=4pF. However, the probe capacity of the observation oscilloscope
13pF is connected in parallel to Cbc + Cbe. In this way, the calculated values and experimental values are in complete agreement, and the refresh operation has been experimentally confirmed.

In the above refresh operation, an example has been described in which the collector is grounded as shown in FIG. 3, but it is also possible to perform the refresh operation with the collector at a positive potential. At this time, even if a refresh pulse is applied to the base-collector junction diode Dbc18, if the positive potential applied to the collector is greater than the potential applied to the base by this refresh pulse, the base-collector junction diode Dbc18 becomes non-active. Since it remains conductive, current flows only through the base-emitter junction diode Dbe16. For this reason, the base potential decreases relatively more slowly than when the collector is grounded, but basically the same high-speed refresh operation as explained above is performed. It is.

In other words, in the relationship between the base potential and the refresh time shown in FIG. 7a, the diagonal straight line when the base potential decreases in FIG. 7a shifts to the right, that is, in a direction that requires more time. Therefore, if you use the same refresh voltage V _RH as when the collector is grounded, it will take time to refresh, but if you slightly increase the refresh voltage V _RH , you can achieve a high-speed refresh just like when the collector is grounded. Operation is possible.

The above is an explanation of the basic operations of the photosensor cell according to the above configuration, which consists of a charge accumulation operation, a readout operation, and a refresh operation by light incidence.

As explained above, the basic structure of the optical sensor cell according to the above configuration is
Publication No. 150878, Japanese Patent Publication No. 56-157073, Japanese Patent Publication No. 157073, Japanese Patent Publication No. 157073
Compared to Publication No. 56-165473, it has an extremely simple structure and is fully compatible with future higher resolutions, as well as its excellent characteristics of low noise, high output, wide dynamic range, and amplification function.
Advantages such as non-destructive readout are preserved as they are.

Next, an embodiment of the photoelectric conversion device of the present invention, which is configured by two-dimensionally arranging the optical sensor cells according to the configuration described above, will be described with reference to the drawings.

FIG. 4 shows a circuit configuration diagram of a photoelectric conversion device in which basic optical sensor cell structures are two-dimensionally arranged in a 3×3 arrangement.

The basic photosensor cell 30 surrounded by the dotted line already described (this time indicates that the collector of the bipolar transistor is connected to the substrate and the substrate electrode); the horizontal line 31 for applying read pulses and refresh pulses; 31', 3
1'', vertical shift register 32 for generating read pulses, buffer MOS between vertical shift register 32 and horizontal lines 31, 31', 31''
Terminal 34 for applying pulses to the gates of transistors 33, 33', 33'', buffer MOS transistors 35, 35', 35'' for applying refresh pulses, and terminals for applying pulses to their gates. 36, a terminal 37 for applying a refresh pulse, a vertical line 38, 3 for reading out the stored voltage from the basic photosensor cell 30;
8', 38'', horizontal shift register 39 that generates pulses for selecting each vertical line, gate MOS transistors 40, 40', 40'' for opening and closing each vertical line, and reading the accumulated voltage to the amplifier section. An output line 41 for outputting the output, and an output line 41 for refreshing the charge accumulated in the output line after reading.
MOS transistor 42, MOS transistor 42
Terminal 4 for applying refresh pulse
3. Bipolar for amplifying the output signal,
Transistors 44 such as MOS, FET, J-FET,
A load resistor 45, a terminal 46 for connecting the transistor and the power supply, an output terminal 47 of the transistor, and vertical lines 40, 40', 4 in the read operation.
0″ to refresh the charge accumulated in
MOS transistors 48, 48', 48'', and
This photoelectric conversion device is constituted by a terminal 49 for applying a pulse to the gates of MOS transistors 48, 48', 48''.

The operation of this photoelectric conversion device will be explained using pulse timing diagrams shown in FIGS. 4 and 8.

In FIG. 8, a section 61 corresponds to a refresh operation, a section 62 corresponds to an accumulation operation, and a section 63 corresponds to a read operation.

At time _t1 , the substrate potential, that is, the collector potential 64 of the photosensor cell portion is kept at the ground potential or positive potential, and FIG. 8 shows that it is kept at the ground potential. Regardless of whether it is at ground potential or positive potential, as already explained, the only difference is the time required for refreshing.
There is no change in the basic operation. The potential 65 of the terminal 49 is
is in a high state, and the MOS transistors 48, 4
8', 48'' are kept conductive and each photosensor cell is grounded through a vertical line 38, 38', 38''. Further, a voltage is applied to the terminal 36 that makes the buffer MOS transistor conductive as shown by a waveform 66, and the buffer MOS transistors 35, 35', and 35'' for refreshing the entire screen at once are in a conductive state.In this state, the terminal When a pulse is applied to 37 as shown in waveform 67, a voltage is applied to the base of each photosensor cell through the horizontal lines 31, 31', 31'', and as explained above, a refresh operation is entered and the previously accumulated data is The charged charges are refreshed according to the complete refresh mode or the transient refresh mode. The pulse width of the waveform 67 determines whether the mode is a complete refresh mode or a transient refresh mode.

At time _t2 , as already explained, the base of the transistor of each photosensor cell is reverse biased with respect to the emitter, and the next accumulation period 62
Move to. In this refresh section 61,
As shown in the figure, all other applied pulses are kept low.

In the accumulation operation section 62, the substrate voltage, that is, the collector potential waveform 64 of the transistor is set to a positive potential. This allows electrons among electron-hole pairs generated by light irradiation to flow quickly toward the collector side. However, keeping the collector potential at a positive potential is not an essential condition because the base is biased in the reverse direction with respect to the emitter, that is, the image is taken with a negative potential, and even if it is kept at ground potential or a slightly negative potential, it is basic. There is no change in the storage behavior.

In the storage operation state, the potential 65 of the gate terminal 49 of the MOS transistors 48, 48', 48''
is kept high, similar to the refresh interval, and each
The MOS transistor remains conductive. Therefore, the emitter of each photosensor cell is placed on the vertical line 3.
8, 38', and 38''. When holes are accumulated in the base due to strong light irradiation and the base becomes saturated, that is, the base potential becomes forward biased with respect to the emitter potential (ground potential). As you get older, the holes become vertical lines 38, 38',
38", where the base potential change stops and becomes clipped. Therefore,
Even if the emitters of the vertically adjacent optical sensor cells are commonly grounded by the vertical lines 38, 38', 38'', the vertical lines 38, 38', 38''
If 38'' is grounded, no blooming phenomenon will occur.

The way to avoid this blooming phenomenon is to
Even if the transistors 48, 48', 48'' are made non-conductive and the vertical lines 38, 38', 38'' are made floating, the substrate potential, that is, the collector potential 6
This can also be achieved by setting 4 at a slightly negative potential so that when the base potential changes toward a positive potential due to accumulation of holes, the potential flows toward the collector side before the emitter.

Following the accumulation section 62, a readout section 63 begins at time _t3 . At this time _t3 , the potential 65 of the gate terminal 49 of the MOS transistors 48, 48', 48'' is set to low, and the horizontal lines 31, 31',
31″ buffer MOS transistors 33, 33′,
The potential 68 of the gate terminal 33'' is set high to bring each MOS transistor into a conductive state.However, it is not an essential condition that the potential 68 of the gate terminal 34 is set high at time _t3 ; It would be better if the time was earlier than that.

At time _t4 , among the outputs of the vertical shift register 32, those connected to the horizontal line 31 have waveform 6.
Since the MOS transistor 33 is in a conductive state at this time, each of the three photosensor cells connected to this horizontal line 31 is read out. This readout operation is as described above, and the signal voltage generated by the signal charge accumulated in the base region of each photosensor cell appears as it is on the vertical lines 38, 38', 38''. As shown in FIG. 6, the pulse width of the pulse voltage from the shift register 32 is set to such a value that the read voltage with respect to the accumulated voltage maintains a sufficiently linear relationship. As shown above, the emitter is adjusted to be forward biased by V _Bias .

Next, at time _t5 , among the outputs of the horizontal shift register 39, the one connected to the vertical line 38
Only the output to the gate of the MOS transistor 40 becomes high as shown in the waveform 70, the MOS transistor 40 becomes conductive, and the output signal passes through the output line 41, enters the output transistor 44, is current amplified, and is output from the output terminal 47. . After the signal is read out in this way, signal charge due to the wiring capacitance remains in the output line 41, so at time _t6 , a pulse as shown in pulse waveform 71 is applied to the gate terminal 43 of the MOS transistor 42. , the MOS transistor 42 is turned on and the output line 41 is grounded to refresh the remaining signal charge. Similarly, the switching MOS transistors 40, 4
0', 40'' are connected in sequence to form vertical lines 38, 3.
8', 38'' is read out. After reading out the signals from each horizontally arranged line of photosensor cells in this way, the signals from the vertical lines 38, 38', 3
Similar to the output line 41, signal charges due to the wiring capacitance of the output line 8'' remain, so the MOS connected to each vertical line 38, 38', 38''
Transistors 48, 48', and 48'' are turned on by having their gate terminals 49 high, as shown by waveform 65, to refresh this residual signal charge.

Next, at time _t8 , among the outputs of the vertical shift register 32, the output connected to the horizontal line 31' becomes high as shown by waveform 69', and the accumulated voltage of each photosensor cell connected to the horizontal line 31' becomes The signals are read out to each vertical line 38, 38', 38''. Thereafter, signals are sequentially read out from the output terminal 47 by the same operation as before.

In the above explanation, the operating conditions applied to application fields where the storage section 62 and the readout section 63 are clearly separated, such as still videos, for which research and development have been actively conducted recently, have been explained. The present invention can also be applied to applications such as cameras where the operation in the storage section 62 and the operation in the readout section 63 are performed simultaneously by changing the pulse timing shown in FIG. 8. However, the refresh function at this time is not a one-time refresh function for the entire screen, but requires a refresh function for each line. For example, after the signals of each photosensor cell connected to the horizontal line 31 are read out,
At time _t7 , MOS transistors 48, 48', 4 are activated to erase the charge remaining on each vertical line.
8'' is made conductive, and at this time a refresh pulse is applied to the horizontal line 31.In other words, in the waveform 69, a pulse with different pulse voltage and pulse width is generated at time _t7 as well as at time _t4 . This can be achieved by using a vertical shift register with the same configuration as the one on the left. It is also possible to achieve this by providing a second vertical shift register on the right side and operating it with a timing shift from that of the vertical register provided on the left side.

At this time, in the accumulation state as described above, the degree of freedom in controlling the blooming by controlling the potentials of the emitter and collector of each photosensor cell is reduced. However, as explained in the basic operation section, in the read state, the base
Since the configuration is such that high-speed reading is possible when a bias voltage of V _Bias is applied, as can be seen from the graph in Figure 5, when V _Bias is not applied,
Due to the saturation of each photosensor cell, vertical lines 28,
The signal charges flowing out to 28' and 28'' are extremely small, and the blooming phenomenon does not pose a problem at all.

Moreover, the photoelectric conversion device according to this embodiment can obtain extremely excellent characteristics with respect to the smear phenomenon. The smear phenomenon is an operational and structural problem that occurs in CCD type imaging devices, especially frame transfer type, in which charge is transferred to the area irradiated with light. This problem occurs because carriers generated deep in the semiconductor due to light are accumulated in the charge transfer section.

Furthermore, in MOS type imaging devices, this problem arises because carriers generated deep in the semiconductor due to long wavelength light are accumulated on the drain side of the switching MOS transistor grounded in each photosensor cell.

On the other hand, in the photoelectric conversion device according to this embodiment, the smear phenomenon that occurs due to its operation and structure does not easily occur, and the phenomenon that carriers generated deep in the semiconductor due to long wavelength light are accumulated does not occur. . However, there is a concern that the electrons and holes generated relatively near the surface of the emitter of the optical sensor cell will accumulate, but this is because the emitter is grounded in the accumulation operation state during the bulk refresh operation. Therefore, electrons are not accumulated,
No smear phenomenon occurs. In addition, during the line refresh operation applied to ordinary television cameras, during the horizontal blanking period, the vertical line is grounded and refreshed before reading out the accumulated voltage on the vertical line, so at the same time the emitter The electrons accumulated during one horizontal scanning period flow out, and therefore almost no smear phenomenon occurs. As described above, in the photoelectric conversion device according to this embodiment, smear development occurs to an essentially negligible extent due to its structure and operation, which is one of the major advantages of the photoelectric conversion device according to this embodiment. It is one.

Further, although the operation of suppressing the blooming phenomenon by manipulating the potentials of the emitter and collector in the storage operation state has been described above, it is also possible to control the γ characteristic using this.

In other words, during the accumulation process, the potential of the emitter or collector is temporarily set to a certain negative potential, and of the carriers accumulated in the base, the holes that are accumulated in a greater number than the number of carriers that give this negative potential are emittered. Alternatively, the data may be flowed to the collector side. As a result, the relationship between the accumulated voltage and the amount of incident light exhibits the γ=1 characteristic of silicon crystal when the amount of incident light is small.
In a place where the amount of incident light is large, γ becomes smaller than 1. In other words, it is possible to provide the characteristic of γ=0.45, which is normally required for television cameras, using polygonal line approximation. If the above operation is performed once during the storage operation, it becomes a one-fold line approximation, and if the negative potential applied to the emitter or collector is changed twice as appropriate, it is also possible to have a two-fold line type γ characteristic.

Further, in the above implementation, the silicon substrate is used as a common collector, but a structure may be adopted in which a buried n ⁺ region is provided like a normal bipolar transistor and the collector is divided for each line.

In addition to the pulse timing shown in FIG. 8, clock pulses for driving the vertical shift register 32 and the horizontal shift register 39 are required for actual operation.

FIG. 9 shows an equivalent circuit related to the output signal.
The capacitance C _V 80 is the wiring capacitance of the vertical lines 38, 38', 38'', and the capacitance C _H 81 is the wiring capacitance of the output line 41.
shows the wiring capacitance of each. The equivalent circuit on the right side of FIG. 9 is in the read state, and the switching MOS transistors 40, 4
0' and 40'' are in a conductive state, and the resistance value in the conductive state is shown by a resistor R _M 82. Also, the amplifying transistor 44 is shown by an equivalent circuit using a resistor r _e 83 and a current source 84. The MOS transistor 42 for refreshing the charge accumulation caused by the wiring capacitance of the output line 41 is non-conductive in the read state and has high impedance, so it is omitted in the equivalent circuit on the right side.

Each parameter of the equivalent circuit is determined by the size of the photoelectric conversion device actually constructed. For example, the capacitance C _V 80 is about 4 pF, and the capacitance C _H 81
is about 4 pF, the conduction state resistance R _M 82 of the MOS transistor is about 3 KΩ, and the current amplification factor β of the bipolar transistor 44 is about 100. As shown in the figure.

In FIG. 10, the horizontal axis represents the time [μs] from the moment the switching MOS transistors 40, 40', 40'' became conductive, and the vertical axis represents the time from the moment when the switching MOS transistors 40, 40', 40'' became conductive.
The output voltage [V] appearing at the output terminal 47 when a signal charge is read from each photosensor cell and a voltage of 1 volt is applied to the wiring capacitance C _V 80 of 8' and 38'' is shown, respectively.

The output signal waveform 85 shows that the load resistance R _E 45 is 10KΩ,
86 is load resistance R _E 45 is 5KΩ, 87 is load resistance
This is when R _E 45 is 2KΩ, and in both cases, the peak value is about 0.5V due to capacitance division between C _V 80 and _CH 81. Naturally, the larger the load resistance R _E 45 is, the smaller the amount of attenuation is, resulting in a desirable output waveform. The rise time is approximately 20nsec when the above parameter values are used.
And it is fast. Resistance R _M of switching MOS transistors 40, 40', 40'' in conduction state
By reducing the wiring capacitance C _V ,
By reducing _CH , even faster reading is possible.

In a photoelectric conversion device using photosensor cells with the above configuration, the voltage appearing at the output is large due to the amplification function of each photosensor cell, so the final stage amplification amplifier is also quite simple compared to a MOS type imaging device. That's fine. In the above example, a one-stage bipolar transistor type was used, but it is of course possible to use other systems, such as a two-stage structure. If a bipolar transistor is used as in this example, the CCD
1/f, which is easily noticeable on images, is generated from the MOS transistor of the final stage amplifier in an imaging device.
The problem of noise does not occur in the photoelectric conversion device of this embodiment, and it is possible to obtain image quality with an extremely good S/N ratio.

As mentioned above, in a photoelectric conversion device using a photosensor cell with the above configuration, the amplification amplifier at the final stage can be extremely simple, so the configuration shown in FIG. 4 in which only one amplification amplifier at the final stage is provided Instead of the type shown in the embodiment shown, it is also possible to install a plurality of amplifiers so that one screen can be divided into a plurality of parts and read out.

FIG. 11 shows an example of a divided readout method. The embodiment shown in FIG. 11 is an example in which the horizontal direction is divided into three parts and three final stage amplifiers are installed. The basic operation is almost the same as that explained using the embodiment shown in FIG. 4 and the timing diagram shown in FIG.
In the embodiment shown in FIG. 11, three equivalent horizontal shift registers 100, 101, and 102 are provided, and when a starting pulse is input to the terminal 103 for applying these starting pulses, the first column and (n+1) column are eye,
The outputs of the sensor cells connected to the (2n+1)th column (n is an integer, and in this embodiment, the number of picture elements in the horizontal direction is 3n) are read out simultaneously. At the next point in time, the second column, (n+2) column,
The (2n+2)th column will be read.

According to this embodiment, when the time to read out one horizontal line is fixed, the horizontal scanning frequency is 1/3 compared to the system with one final stage amplifier. A high-speed analog-to-digital converter is unnecessary for applications where the frequency is sufficient, the horizontal shift register is simple, and the output signal from a photoelectric conversion device is converted into analog-to-digital for signal processing. This is a major advantage of this method.

In the embodiment shown in FIG. 11, three equivalent horizontal shift registers are provided, but the same function can be provided with only one horizontal shift register. . The first example of this case is
Shown in Figure 2.

The embodiment of FIG. 12 uses the horizontal switching MOS transistor of the embodiment shown in FIG.
Only the middle part of the final stage amplifier is shown, and the other parts are omitted because they are the same as the embodiment shown in FIG.

In this example, one horizontal shift register 1
The output from 04 is in the 1st column, (n+1) column,
It is connected to the gate of the switching MOS transistor in the (2n+1)th column so that those lines can be read out simultaneously. At the next point in time, the second column, (n
The +2)th column and the (2n+2)th column are read out.

According to this embodiment, although the number of wirings to the gate of each switching MOS transistor increases,
It is possible to operate with only one horizontal shift register.

In the examples shown in FIGS. 11 and 12, three output amplifiers are provided, but it goes without saying that this number may be increased depending on the purpose.

In both the embodiments shown in FIGS. 11 and 12, the starting pulses and clock pulses for the horizontal shift register and vertical shift register are omitted, but they are provided in the same chip like other refresh pulses. clock pulse generator or
It is supplied from a clock pulse generator on another chip.

In this split readout method, when refreshing horizontal lines or all screens at once, the storage time is slightly different between the n-th and (n+1)-th row photosensor cells, which causes a slight difference in the dark current component and signal component. It is conceivable that discontinuity may occur and be noticeable on the image, but the amount of this is small and poses no practical problem. Also, this is
Even if it exceeds the allowable limit, using an external circuit to correct it is a conventional method that generates a square wave, subtracts this from the dark current component, and multiplies and divides this by the signal component. This is easily possible by using correction techniques.

When capturing a color image using such a photoelectric conversion device, a stripe filter or a mosaic filter is placed on-chip on top of the photoelectric conversion device, or a separately manufactured color filter is attached to the photoelectric conversion device. It is possible to get a signal.

As an example, when R, G, and B stripe filters are used, a photoelectric conversion device using a photosensor cell according to the above configuration can obtain the R signal, G signal, and B signal from separate final stage amplifiers. It is possible. An example of this is shown in FIG. Like FIG. 12, FIG. 13 also shows only the area around the horizontal shift register. The rest is the same as Fig. 4 and Fig. 11, except that the first column is an R color filter, the second column is a G color filter,
It is assumed that color filters are installed in the third column, such as the B color filter and the fourth column, the R color filter. As shown in Figure 13, the first row,
Each vertical line in the 4th column, 7th column, etc. is output line 1
10, which takes out the R signal. Also 2
Each vertical line of the 5th column, 8th column, etc. is connected to an output line 111, which takes out the G signal. Similarly, each vertical line of the 3rd column, 6th column, 9th column, etc. is connected to the output line 112.
Take out the signal. Output lines 110, 111, 1
12 are each formed on-chip or connected to a refresh MOS transistor and a final stage amplifier, for example, an emitter follower type bipolar transistor, and each color signal is output separately.

FIG. 14 shows a diagram for explaining the basic structure and operation of another example of a photosensor cell constituting a photoelectric conversion device according to another example of the present invention. In addition, the equivalent circuit and the overall circuit configuration diagram are shown in Figure 15a.
Shown below.

The optical sensor cell shown in FIG. 14 is an optical sensor cell that can simultaneously perform a read operation and a line refresh using the same horizontal scan pulse. In Figure 14, the first
The difference from the configuration shown in the figure is that in the case of FIG. 1, there was only one MOS capacitor electrode 9 connected to the horizontal line wiring 10, but there are also MOS capacitor electrodes 120 connected to the vertically adjacent optical sensor cells.
are connected and viewed from one optical sensor cell,
In addition, the emitters 7 and 7' of the vertically adjacent photosensor cells in the figure are double-layered wiring 8 and wiring 121 (in Figure 14, the vertical lines are reduced to one). As you can see, the two lines are arranged through an insulating layer), that is, the emitter 7 is connected to the wiring 8 through the contact marl 19.
Another difference is that the emitters 7' are connected to the interconnections 121 through contact holes 19'.

This becomes clearer when looking at the equivalent circuit shown in FIG. 15a. That is, the MOS capacitor 150 connected to the base of the optical sensor cell 152 is connected to the horizontal line 31, and the MOS capacitor 151 is connected to the horizontal line 31'. Furthermore, in the diagram of the optical sensor cell 152, the MOS capacitors 150' of the adjacent optical sensor cells 152' below are connected to a common horizontal line 31'.

The emitter of the optical sensor cell 152 is on the vertical line 3.
8, the emitters of the photosensor cells 152' are connected alternately to the vertical line 138, and the emitters of the photosensor cell 152'' are connected to the vertical line 38, respectively.

The equivalent circuit of FIG. 15a differs from the imaging device of FIG. 4, other than the basic photosensor cell section described above, in that it includes a switching MOS transistor 48 for refreshing the vertical line 38 and In addition to the switching MOS transistor 148 for refreshing and the switching MOS transistor 40 for selecting the vertical line 38, a switching MOS transistor 140 for selecting the vertical line 138 is added, and one output amplifier system is added. The configuration of this output system consists of 4 switching MOS transistors for refreshing each line.
8 and 148 are connected, and a switching MOS for horizontal scanning is also installed.
It is also possible to use only one output amplifier as shown in FIG. 15b using transistors. FIG. 15b shows only the vertical line selection and output amplifier system portions of FIG. 15a.

This optical sensor cell in FIG. 14 and FIG. 15a
According to the embodiment shown in , the following operations are possible. That is, when the reading operation of each optical sensor cell connected to the horizontal line 31 is completed and the horizontal blanking period in the TV operation is in progress, the output pulse from the vertical shift register 32 is outputted to the horizontal line 31'. The optical sensor cell 152 that has been read out is refreshed through the capacitor 151. At this time, the switching MOS
Transistor 48 is rendered conductive and vertical line 38 is grounded.

Further, the output of the photosensor cell 152' is read out to the vertical line 138 through the MOS capacitor 150' connected to the horizontal line 31'. At this time, naturally, the switching MOS transistor 148 is rendered non-conductive, and the vertical line 138
is in a floating state. In this way, with one vertical scan pulse, it is possible to simultaneously refresh the optical sensor cells that have already been read out and read out the optical sensor cells of the next line using the same pulse. At this time, as already explained, the voltage at the time of refreshing and the voltage at the time of reading are different because a bias voltage is applied at the time of reading due to the necessity of high-speed reading. This is achieved by changing the areas of the electrode 9 and the MOS capacitor electrode 120 so that even if the same voltage is applied to each electrode, different voltages are applied to the base of each photosensor cell.

That is, the area of the refresh MOS capacitor is smaller than the area of the read MOS capacitor. As in this example, when refreshing one line at a time instead of refreshing all sensor cells at once, the first
Although the collector may be formed of an n-type or n-substrate as shown in FIG. b, it may be desirable to separate the collectors for each horizontal line. When the collector is a substrate, the collectors of all the photosensor cells are a common area, so that a constant bias voltage is applied to the collectors in the storage and light reception/readout states. Of course, as explained above, floating-based refresh can be performed between the emitters even when a bias voltage is applied to the collector. However, in this case, there is a drawback that at the same time that the base region is refreshed, a wasteful current flows between the emitter collector of the cell to which the refresh pulse is applied, increasing power consumption.
In order to overcome these drawbacks, instead of making the collectors of all the sensor cells a common area, the collectors of the sensor cells arranged in each horizontal line are common, but the collectors of each horizontal line are structured to be separated from each other. That is, in connection with the structure shown in FIG. 1, the substrate is of a p-type, and the p-type substrate has a structure in which n ⁺ buried regions separated from each other are provided for each horizontal line of the collector. The n ⁺ buried regions of adjacent horizontal lines may be separated by a structure in which a p region is interposed therebetween. Dielectric isolation is better for reducing collector capacitance buried along horizontal lines. In FIG. 1, since the collector is comprised of a substrate, all isolation regions surrounding the sensor cell are provided to approximately the same depth. On the other hand, in order to separate the collectors of each horizontal line from each other, the separation area in the horizontal line direction is made deeper than the separation area in the vertical line direction by a necessary value.

If the collector is separated for each horizontal line, if the voltage of the collector of that horizontal line is grounded when reading is finished and the refresh operation starts, the emitter-collector current as described above will not flow, and the consumption will be reduced. Does not result in an increase in power. When the refresh ends and the charge storage operation starts based on the optical signal, a predetermined bias voltage is applied to the collector region again.

According to the equivalent circuit shown in FIG. 15a, outputs are alternately output to the output terminals 47 and 147 for each horizontal line. As already explained, it is also possible to take out the output from one amplifier by using the configuration as shown in FIG. 15b.

As explained above, according to this embodiment, line refresh is possible with a relatively simple configuration.
It can also be applied to fields of application such as ordinary television cameras.

As another embodiment of the present invention, an optical sensor cell may be provided with a plurality of emitters, or one emitter may be provided with a plurality of contacts.
A type that takes out multiple outputs from one optical sensor cell is considered.

This is because each optical sensor cell of the photoelectric conversion device according to the present invention has an amplification function, so even if multiple wiring capacitors are connected to each optical sensor cell in order to extract multiple outputs from one optical sensor cell, the optical sensor cell This is due to the fact that the internally generated accumulated voltage Vp can be read out to each output without attenuating too much.

In this way, by having a configuration in which multiple outputs can be taken out from each optical sensor cell, many advantages can be added to the photoelectric conversion device formed by arranging a large number of each optical sensor cell in terms of signal processing, noise countermeasures, etc. is possible.

Next, an example of a method for manufacturing a photoelectric conversion device according to the present invention will be described. Figure 16 shows selective epitaxial growth (N. Endo et al, “Novel device isolation
technology with selected epihaxial growth”
Tech.Dig.of 1982 IEDM, pp. 241-244)
An example of the manufacturing method using .

An n ⁺ region 11 for contact is formed on the back side of an n-type Si substrate 1 with an impurity concentration of about 1 to 10 × 10 ¹⁶ cm ^-3 .
Provided by diffusion of As or P. In order to prevent autodoping from the n ⁺ region, an oxide film and a nitride film (not shown) are usually provided on the back surface.

The substrate 1 used has impurity concentration and oxygen concentration controlled to be uniform. That is, a crystal wafer having a sufficiently long and uniform carrier line time is used. For example, such a thing
Crystals produced by the MCZ method are suitable. An oxide film of about 1 μm is formed on the surface of the substrate 1 by wet oxidation. That is, H ₂ O atmosphere or (H ₂
+O ₂ ) oxidizes in the atmosphere. High-pressure oxidation at a temperature of about 900°C is suitable for obtaining a good oxide film without producing stacking faults.

On top of that, for example, a layer with a thickness of about 2 to 4 μm is added.
Deposit the SiO ₂ film by CVD. (N ₂ + SiH ₄ + O ₂ ) to the desired thickness at a temperature of about 300 to 500℃ using a gas system.
Deposit the _SiO2 film. The molar ratio of O ₂ /SiH ₄ is set to about 4 to 40, although it depends on the temperature. Through the photolithography process, the oxide film in other areas is made into (CF ₄ +
H ₂ ), C ₂ F ₄ , CH ₂ F _2, etc., using reactive ion etching (step a in Figure 16). For example, when one pixel is provided in 10 x 10 μm ² , Leave the SiOB film in the form of a mesh with a pitch of 10 μm. The width of the SiO ₂ film is selected to be approximately 2 μm, for example.
After removing the damaged layer and contaminant layer on the surface caused by reactive ion etching by Ar/Cl ₂ gas plasma etching or wet etching, the atmosphere is sufficiently cleaned by evaporation in an ultra-high vacuum or by using a load lock method. Amorphous silicon 301 is deposited using sputtering or low pressure CVD in which SiH ₄ gas is irradiated with a CO ₂ laser beam (step b in FIG. 16).
By anisotropic etching by reactive ion etching using gases such as CBrF ₃ , CCl ₂ F ₂ , Cl _{2 ,} etc.
Remove the amorphous silicon other than that deposited on the side surface of the SiO ₂ layer (step c in Figure 16). After sufficiently removing the damaged layer and contamination layer as before,
Thoroughly clean the silicon substrate surface and remove (H ₂ +
The silicon layer is selectively grown using a gas system (SiH ₂ , Cl ₂ +HCl). Growth is performed under reduced pressure of several tens of Torr, the substrate temperature is set at 900 to 1000°C, and the molar ratio of HCl is set to a value higher than a certain level. If the amount of HCl is too small, selective growth will not occur. A silicon crystal layer grows on the silicon substrate, but since the silicon on the SiO ₂ layer is etched by HCl, no silicon is deposited on the SiO ₂ layer (FIG. 16d). The thickness of the n ^- layer 5 is, for example, about 3 to 5 μm. Impurity concentration is preferably between 10 ¹² and 10 ¹⁶ cm
Set it to about ^-3 . Of course, it is possible to deviate from ^{this range} , but the impurity concentration and It is preferable to choose the thickness.

Since commonly available HCl gas contains a large amount of water, an oxide film is constantly formed on the surface of the silicon substrate, making it impossible to expect high-quality epitaxial growth. watery
When HCl is in the cylinder, it reacts with the cylinder material and contains a large amount of heavy metals, mainly iron, which tends to result in an epitaxial layer with heavy metal contamination. Since it is desirable for the epitaxial layer used in the optical sensor cell to have as little dark current component as possible, it is necessary to suppress contamination by heavy metals to the utmost. In addition to using ultra-high purity materials for SiH ₂ Cl ₂ , HCl must also have a particularly low moisture content, preferably at least a low moisture content.
Use 0.5ppm or less. Of course, the lower the water content, the better. For even higher quality epitaxially grown layers, the substrate is first heated to 1150-1250
Oxygen is removed from near the surface by high temperature treatment at around 800°C, followed by long-term heat treatment at around 800°C to generate many micro defects inside the substrate, making it possible to perform intensive getttering with a denuded zone. It is also extremely effective to use it as a substrate. Since epitaxial growth is performed in the presence of the SiO ₂ layer 4 as a separation region, it is desirable that the growth temperature be as low as possible in order to reduce the amount of oxygen taken in from the SiO ₂ . With the commonly used high-frequency heating method, there is a lot of contamination from the carbon susceptor, making it difficult to lower the temperature further.
The wafer direct heating method using lamp heating, which does not involve bringing a carbon susceptor into the reaction chamber, provides a clean growth atmosphere and allows high-quality epitaxial layers to be grown at low temperatures.

Ultra-high purity fused sapphire, which has a lower vapor pressure, is suitable for the wafer support in the reaction chamber. The wafer direct heating method using lamp heating allows for easy preheating of the raw material gas and makes it easy to uniformize the temperature within the wafer surface even when a large flow of gas is flowing.In other words, the wafer direct heating method using lamp heating generates almost no thermal stress. suitable for obtaining. UV irradiation of the wafer surface during growth further improves the quality of the epitaxial layer.

Amorphous silicon is deposited on the sidewalls of the SiO ₂ layer forming the isolation region 4 (step c in FIG. 16). Since amorphous silicon is easily formed into a single crystal by solid phase growth, the crystal near the interface with the SiO ₂ separation region 4 becomes very good. High resistance n ^- layer 5
After forming by selective epitaxial growth (step d in Fig. 16), a P region 6 with a surface concentration of about 1 to 20 × 10 ¹⁶ cm ^-3 is formed by diffusion from doped oxide or by low-dose ion implantation. The layer is formed to a predetermined depth by diffusion using the layer as a source. The depth of p region 6 is, for example, about 0.6 to 1 μm.

The thickness and impurity concentration of p region 6 are determined based on the following considerations. In order to increase the sensitivity, it is desirable to lower the impurity concentration in the p region 6 to reduce Cbe. Cbe is approximately given as follows.

Cbe=Aeε(q・N _A /2εVbi) ^1/2 However, Vbi is the emitter-base diffusion potential and is given by Vbi=kT/q1nN _D N _A /n _i ² . Here, ε is the dielectric constant of the silicon crystal, N _D is the impurity concentration of the emitter, N _A is the impurity density of the portion of the base adjacent to the emitter, and n _i is the intrinsic carrier concentration. The smaller N _A is, the more Cbe
As N A becomes smaller, the sensitivity increases, but if N _A is made too small, the base region will be completely depleted in the operating state, resulting in a punch-through state, so it cannot be made too low. It is set to such an extent that the base region is not completely depleted and a punch-through state occurs.

Thereafter, a thermal oxide film 3 having a thickness of several tens of angstroms to several hundreds of angstroms is formed on the surface of the silicon substrate by (H ₂ +O ₂ ) gas-based steam oxidation at a temperature of approximately 800 to 900°C. On top of that, (SiH ₄ +NH ₃ ) system gas
A nitride film (Si ₃ N ₄ ) 302 is formed to a thickness of about 500 to 1500 Å by CVD. The formation temperature is about 700-900℃. NH ₃ gas, along with HCl gas, also contains a large amount of water in commonly available products. If NH ₃ gas with a high moisture content is used as a raw material, the result will be a nitride film with a high oxygen concentration, resulting in poor reproducibility and an inability to obtain a selective etching ratio with the SiO ₂ film. _NH3 gas also
The moisture content should be at least 0.5 ppm or less. It goes without saying that the lower the water content, the more desirable it is. A PSG film 3 is further formed on the nitride film 302.
00 is deposited by CVD. For the gas system, for example, (N ₂ + SiH ₄ + O ₂ + PH ₃ ) is used,
With a thickness of about 2000 to 3000 Å at a temperature of about 450℃
A PSG film is deposited by CVD (step e in Figure 16). By a photolithography process including two mask alignment processes, an As-doped polysilicon film 304 is deposited on the n ⁺ region 7 and on the refresh and read pulse application electrodes. In this case, a p-doped polysilicon film may be used. For example, by two photolithography steps,
The PSG film, Si ₃ N ₄ film, and SiO ₂ film were all removed from the emitter, leaving the underlying SiO ₂ film in the area where the refresh and readout pulse application electrodes were to be provided.
Only the PSG film and Si ₃ N ₄ film are etched. after that,
As-doped polysilicon (N ₂ + SiH ₄ +
AsH ₃ ) or (H ₂ +SiH ₄ +AsH ₃ ) gas
Deposited by CVD method. Deposition temperature is 550℃~700℃
The film thickness is about 1000 to 2000 Å. Of course, non-doped polysilicon may be deposited by the CVD method and then As or P may be diffused. After the mask alignment photolithography process, the polysilicon film in other parts except on the emitter, refresh and readout pulse application electrodes is removed by etching. When the PSG film is further etched, the polysilicon deposited on the PSG film is removed in a self-aligned manner due to lift-off (first step).
Step f) in Figure 6. Etching of polysilicon film
Etching is performed using a gas system such as C ₂ Cl ₂ F ₄ or (CBrF ₃ +Cl ₂ ), and the Si ₃ N ₄ film is etched using a gas such as CH ₂ F ₂ .

Next, a PSG film 305 is deposited by the gas-based CVD method as described above, and then a contact hole is formed on the polysilicon film for the refresh pulse and read pulse electrodes by a mask alignment process and an etching process. In this state, A1,
Depositing metals such as A1-Si and A1-Cu-Si by vacuum evaporation or sputtering, or using plasma using (CH ₃ ) ₃ Al or AlCl ₃ as the raw material gas.
CVD method or Al-C of the above raw material gas
Al is deposited using a light irradiation CVD method in which the bond or Al-Cl bond is cut by direct light irradiation.
When carrying out the above-mentioned CVD method using (CH ₃ ) ₃ Al or AlCl ₃ as the raw material gas, a large excess of hydrogen is allowed to flow. For narrow and steep contact holes
An excellent method for depositing Al is the CVD method, which raises the substrate temperature to a film thickness of 300 to 400°C in a clean atmosphere that does not contain moisture or oxygen. After completing the patterning of the metal wiring 10 shown in FIG.
An interlayer insulating film 306 is deposited by CVD. 306
is the aforementioned PSG film or CVD SiO ₂ film,
Alternatively, if it is necessary to take water resistance into consideration, an Si ₃ N ₄ film formed by a (SiH ₄ +NH ₃ ) gas-based plasma CVD method is used. In order to keep the hydrogen content in the Si ₃ N ₄ film low, (SiH ₄ + N ₂ )
Uses gas-based plasma CVD method.

In order to increase the electrical withstand voltage of the Si ₃ N ₄ film formed by reducing the damage caused by the plasma CVD method and to reduce the leakage current, the photo CVD method is used.
The Si ₃ N ₄ film is excellent. There are two methods for optical CVD. (SiH ₄ + NH ₃ + Hg) gas system with 2537 Å ultraviolet rays from a mercury lamp from the outside, and (SiH ₄ + NH 3 + Hg) ₃ gas system with 1849 Å ultraviolet rays from a mercury lamp.
This is a method of irradiating UV rays. In both cases, the substrate temperature is about 150 to 350°C.

Through the mask alignment process and etching process,
An insulating film 305,
After drilling a contact hole through 306 using reactive ion etching, Al,
Deposit metals such as Al-Si, Al-Cu-Si, etc. In this case, since the aspect ratio of the contact hole is large, deposition by CVD is superior. After patterning the metal wiring 8 in Fig. 1, a final passivation film is formed.
A Si ₃ N ₄ film or a PSG film 2 is deposited by the CVD method (FIG. 16g).

In this case as well, the film produced by the photo-CVD method is superior. 12 is a metal electrode made of Al, Al-Si, etc. on the back surface.

The method for manufacturing the photoelectric conversion device of the present invention involves a wide variety of steps, and FIG. 16 shows only one example.

The important point of the photoelectric conversion device of the present invention is that the p region 6
The problem lies in how to suppress leakage current between the and n ^- regions 5 and between the p region 6 and the n ⁺ region 7. n ^-
It goes without saying that dark current can be reduced by improving the quality of the region 5, but the problem lies in the interface between the isolation region 4 made of an oxide film or the like and the n ^- region 5. In FIG. 16, for this purpose, the separation area 4 is
We have explained a method in which amorphous Si is deposited on the sidewalls of the substrate and then epitaxially grown. In this case, amorphous Si is made into a single crystal by solid phase growth from the substrate Si during epitaxial growth. Epitaxial growth is performed at a relatively high temperature of about 850°C to 1000°C. Therefore, the board
Before the amorphous Si becomes single crystallized by solid phase growth from Si, microcrystals often begin to grow in the amorphous Si, which causes poor crystallinity. The lower the temperature, the faster the solid-phase growth becomes relatively faster than the speed at which microcrystals begin to grow in amorphous Si. Therefore, before performing selective epitaxial growth, it is necessary to If amorphous Si is made into a single crystal through processing,
The interfacial properties are improved. At this time, if there is a layer such as an oxide film between the substrate Si and the amorphous Si, the start of solid phase growth will be delayed, so an ultra-high cleanliness process is required to prevent such a layer from being included at the boundary between the two.

In addition to the above-mentioned furnace growth, for solid-phase growth of amorphous Si, rapid annealing technology is also effective, for example, by keeping the substrate at a certain temperature and heating it with a fuselage lamp or an infrared lamp, for a period of several seconds to several tens of seconds. When using such techniques, the Si deposited on the sidewalls of the SiO ₂ layer may be polycrystalline. However, polycrystalline Si needs to be deposited using a very clean process and does not contain oxygen, carbon, etc. at the grain boundaries of the polycrystalline material.

After the Si on these SiO ₂ sides is single-crystalized, the Si
This will result in selective growth.

When leakage current at the interface between the SiO ₂ isolation region 4 and the high resistance n ^- region 5 becomes a problem, the high resistance n ^-
Only the part of region 5 adjacent to SiO ₂ separation region 4,
This leakage current problem can be avoided by increasing the n-type impurity concentration. For example, only the region with a thickness of about 0.3 to 1 μm of the n ^- region 5 that contacts the isolated SiO ₂ region 4, for example, about 1 to 10 × 10 ¹⁶ cm ^-3
This increases the impurity concentration in the shape. This configuration can be formed relatively easily. After forming a thermal oxide film with a thickness of about 1 μm on the substrate 1, it is deposited thereon by CVD. First, the SiO ₂ film is made into a SiO ₂ film having a required thickness and containing a predetermined amount of P. Furthermore, a separation region 4 is created by depositing SiO ₂ thereon by CVD. In the subsequent high-temperature process, phosphorus is diffused from the phosphorus-containing SiO ₂ film that exists in the form of a sandwich in the separation region 4 into the high-resistance n ^- region 5, resulting in a good impurity distribution with a high impurity concentration even at the interface. make.

In other words, the structure is as shown in FIG. 17. The isolation region 4 has a three-layer structure, in which 308 is a thermal oxide film SiO ₂ , 309 is a CVD SiO ₂ film containing phosphorus, and 301 is a CVD SiO ₂ film. Adjacent to the separation region 4 and between the n ^- region 5 and the n-region 307, an SiO ₂ film 309 containing phosphorus is formed.
Formed by diffusion from 307 is formed all around the cell. With this structure, the base
Although the collector-collector capacitance Cbc increases, the base-collector leakage current decreases dramatically.

In FIG. 16, an example was explained in which the isolation insulating region 4 was formed in advance and selective epitaxial growth was performed.
After epitaxial growth of the n ^-layer ,
U-group separation technology (A.Hayasaka
et al, “U-groove isolation technique for
high speed bipolar VLSI′S″，Tech.Dig.of
(see IEDM.P.62, 1982).

A photoelectric conversion device according to the present invention forms a bipolar transistor in which a base region, most of which is adjacent to the semiconductor wafer surface, is in a floating state in a region surrounded by an isolation region made of an insulator, This is a device that photoelectrically converts optical information by controlling the potential of a floating base region with an electrode provided on a part of the base region via a thin insulating layer. An emitter region made of a high impurity concentration region is provided in a part of the base region, and this emitter is connected to a MOS transistor operated by a horizontal scan pulse. The aforementioned electrode provided on a portion of the floating base region via a thin insulating layer is connected to a horizontal line. The collector provided inside the wafer may be composed of a substrate, or depending on the purpose, it may be composed of high-concentration impurity embedded regions separated for each horizontal line on a high-resistance substrate of the opposite conductivity type. There is also. The pulse voltage applied when reading out a signal is substantially larger than the pulse voltage when refreshing the floating base region using the electrode provided through the insulating layer. In fact, you may use a pulse train with two types of voltages, or as explained in the double capacitor structure,
The capacitance C _px of the read MOS capacitor electrode may be made larger than the capacitance C _px of the MOS capacitor electrode. By applying a refresh pulse, optically excited carriers are accumulated in the floating base region, which is brought into a reverse bias state, and a signal based on the optical signal is stored. When reading out the signal, the base-emitter is deeply biased in the forward direction. The feature is that a readout pulse voltage is applied so that signals can be read out at high speed. As long as it has these characteristics, the photoelectric conversion device of the present invention may be realized in any structure, and it is needless to say that it is not limited to the structure described in the above embodiments.

For example, the same applies to a structure in which the conductivity type is completely reversed from that described in the above embodiment. However, at this time, it is necessary to completely reverse the polarity of the applied voltage. In a structure in which the conductivity types are completely reversed, the region becomes n-type. That is, the impurities constituting the base are As and P. When the surface of a region containing As or P is oxidized, As or P becomes Si/
Pile up on the Si side of the SiO ₂ interface. That is,
A strong drift electric field is generated inside the base from the surface to the inside, and the photoexcited holes immediately escape from the base to the collector side, and electrons are efficiently accumulated in the base.

If the base is p-type, the commonly used impurity is boron. When the surface of the p-region containing boron is thermally oxidized, boron is incorporated into the oxide film, so the boron concentration in the Si near the Si/SiO ₂ interface becomes slightly lower than the boron concentration inside. This depth depends on the oxide film thickness, but is usually several hundred angstroms.
A reverse drift electric field for electrons is generated near this interface, and electrons photoexcited in this region tend to be collected on the surface. If this continues, the region where this reverse drift electric field is generated will become a dead region, but since the photoelectric conversion device of the present invention has an n ⁺ region along a part of the surface, the p-region Si/ Electrons gathered at the SiO ₂ interface flow into this n ⁺ region before being recombined. Therefore, even if there is a region where, for example, boron is reduced near the Si/SiO ₂ interface and a reverse drift electric field occurs, it hardly becomes a dead region. Rather, if such a region exists at the Si/SiO ₂ interface, the accumulated holes will be transferred to the Si/SiO ₂
By separating the holes from the interface and making them exist inside, the effect of holes disappearing at the interface is eliminated, and the effect of accumulating holes at the base of the p-layer becomes good, which is extremely desirable.

As described above, the photoelectric conversion device of the present invention accumulates carriers excited by light in the base region, which is the control electrode region in a floating state. i.e. Base Store Image
It is a device that should be called a Sensor, and is abbreviated as BASIS.

In addition to the solid-state imaging device described above, the photoelectric conversion device according to the present invention can also be used, for example, an image input device,
It can also be applied to image input devices such as facsimiles, workstations, digital copying machines, and word processors, OCR, barcode reading devices, and photoelectric conversion object detection devices for autofocus of cameras, video cameras, 8 mm cameras, etc.

[Effects of the Invention] According to the present invention, the potential of the control electrode region as the third semiconductor region is controlled independently of the main electrode region as the first and fourth semiconductor regions, so that the output voltage signal is High-speed readout is possible while ensuring good linearity.

Furthermore, during storage, the saturation voltage can be determined by the voltage that reverse biases the third semiconductor region with respect to the fourth semiconductor region, and during refresh, the third semiconductor region can be biased deeply in the forward direction with respect to the fourth semiconductor region. It operates at high speed, has a wide dynamic range, and can provide a good output signal with no afterimage or noise. Furthermore, blue sensitivity characteristics are improved.

Since the photoelectric conversion device of the present invention can configure one pixel with one transistor, it is extremely easy to increase the density, and at the same time, due to its structure, there is little blooming and smear, and it has high sensitivity, and its dynamic range is wide. Since it has an internal amplification function, it generates a large signal voltage regardless of the wiring capacity, so it has the characteristics of low recording and easy peripheral circuitry. For example, its industrial value is extremely high as a future high-quality solid-state imaging device.

[Brief explanation of the drawing]

1 to 3 and 5 to 7 are diagrams for explaining the main structure and basic operation of an optical sensor cell according to an embodiment of the present invention. Figure 1a is a plan view, b is a sectional view, c is an equivalent circuit diagram, Figure 2 is an equivalent circuit diagram during read operation, and Figure 3 is an equivalent circuit diagram.
The figure is an equivalent circuit diagram during refresh operation, and Figure 5 is a graph showing the relationship between readout time and readout voltage.
Figure 6a shows the relationship between storage voltage and readout time, Figure 6b shows the relationship between bias voltage and readout time, and Figures 7a to c show the relationship between refresh time and base potential. It is a graph.
FIG. 4 is a circuit diagram of a photoelectric conversion device constructed from the optical sensor shown in FIG. 1. 8 to 10 are explanatory diagrams of the photoelectric conversion device of FIG. 4 configured by the optical sensor shown in FIG. 1, and FIG. 8 is a pulse timing diagram. FIG. 9 is an equivalent circuit diagram related to the output signal, and FIG. 10 is a graph showing the output voltage from the moment of conduction in relation to time.
11, 12 and 13 are circuit diagrams showing other photoelectric conversion devices. FIG. 14 is a plan view for explaining the main structure of a modified example of the present invention. 15th
This figure is a circuit configuration diagram of a photoelectric conversion device constructed from the optical sensor cell shown in FIG. 14. FIGS. 16 and 17 are cross-sectional views showing an example of a method for manufacturing a photoelectric conversion device of the present invention. DESCRIPTION OF SYMBOLS 1... Silicon substrate, 2... PSG film, 3... Insulating oxide film, 4... Element isolation region, 5... N ^- region (collector region), 6... P region (base region), 7, 7'...
n ⁺ region (emitter region), 8...wiring, 9...electrode,
DESCRIPTION OF SYMBOLS 10... Wiring, 11... n ⁺ area, 12... Electrode, 13
...capacitor, 14...bipolar transistor,
15, 17... Junction capacitance, 16, 18... Diode, 19, 19'... Contact portion, 20... Light, 2
8... Vertical line, 30... Photo sensor cell, 31... Horizontal line, 32... Vertical shift register, 33,3
5...MOS transistor, 36, 37...terminal, 3
8...Vertical line, 39...Horizontal shift register, 4
0...MOS transistor, 41...output line, 4
2...MOS transistor, 43...terminal, 44...transistor, 45...load resistance, 46...terminal, 47
...terminal, 48...MOS transistor, 49...terminal,
61, 62, 63... section, 64... collector potential,
67... Waveform, 80, 81... Capacitance, 82, 83... Resistor, 84... Current source, 100, 101, 102... Horizontal shift register, 111, 112... Output line, 138... Vertical line, 140... MOS transistor, 148... MOS transistor, 150,
150'...MOS capacitor, 152, 152'...
Optical sensor cell, 300...amorphous silicon,
302...Nitride film, 303...PSG film, 304...Polysilicon, 305...PSG film, 306...Interlayer insulating film.

Claims (1)

[Scope of Claims] 1: a first semiconductor region of a first conductivity type electrically connected to a first electrode; a first semiconductor region disposed adjacent to the first semiconductor region and higher than the first semiconductor region; a second semiconductor region having a resistivity; and a second conductivity type different from the first conductivity type, which is disposed adjacent to the second semiconductor region and is disposed with respect to the second electrode via an insulating region. a fourth semiconductor region of a first conductivity type, which is disposed adjacent to the third semiconductor region and is electrically connected to a third electrode connected to an output circuit including a capacitive load; A photoelectric conversion method using a photoelectric conversion device including a transistor having the following, wherein the fourth semiconductor region is held at a ground potential or in a floating state, and a junction between the fourth semiconductor region and the third semiconductor region is an accumulation step of biasing in the opposite direction and irradiating the transistor with light in order to accumulate carriers generated by receiving optical energy in the third semiconductor region; a readout step comprising biasing a junction between the third semiconductor region and the fourth semiconductor region in the forward direction and reading out an output signal corresponding to the accumulated carrier from the third electrode as a voltage at the capacitive load; A refreshing step including holding the semiconductor region at a ground potential and biasing the junction between the third semiconductor region and the fourth semiconductor region in the forward direction to remove accumulated carriers. photoelectric conversion method.