JP2010258711A - Near-infrared image sensor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a small-sized near-infrared sensor for obtaining a high S/N signal or a sharp image even if there is a dark current. <P>SOLUTION: A near-infrared image sensor includes: one or more photodetectors each formed on an InP substrate 1 and having a photodetection layer 3; a modulation section 50 positioned at an incident side of near-infrared light; and a signal processing section 70 positioned at a back side of near-infrared light. A band gap wavelength of the photodetection layer 3 is equal to or more than 1.2 μm and less than or equal to 3 μm, the modulation section 50 is formed from MEMS with silicon as a main component, covers one or more photodetectors and is integrated with the photodetectors, and the signal processing section 70 then includes a signal readout circuit for reading out signals of the photodetectors and a signal detector for detecting signals from the photodetectors. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a near-infrared image sensor for signal light in the near-infrared region.

Group III-V compound semiconductors are attracting attention as compound semiconductors having a band gap energy corresponding to the near infrared wavelength range of about 1.7 μm or longer wavelengths of about 2 to 3 μm, and research and development are in progress. . For example, a light receiving element having a light receiving layer of a type II multiple quantum well structure in which (InGaAs / GaAsSb) is paired on InP has been proposed (Non-Patent Document 1). In addition, single-layer GaInNAs has been proposed as a compound semiconductor having light receiving sensitivity up to a wavelength of about 2.5 μm (Patent Document 1).
However, even if a near-infrared sensor is constituted by the light receiving element, a clear image cannot be obtained. One of the major reasons is that the dark current in the light receiving element is large. In the prior art, an image is obtained by alternately measuring a difference between when there is no light input and when there is a light input and holding the measurement value. If the dark current, which is one of the major causes of noise, is large, the light input level is almost occupied by the dark current when there is no light input. If there is light input but the input level is small, an image is obtained. The output (current or voltage) of each pixel is buried in the output caused by the dark current, and a clear image cannot be obtained.
In order to overcome the problem of noise including dark current, an apparatus using an optical chopper has been disclosed for diagnosing the presence or absence of mastitis using near infrared light or the like (Patent Document 2). In addition, a medical photometer has been proposed in which noise of a signal is reduced while always measuring a dark signal (Patent Document 3). In addition, a sensor signal processing method has been proposed in which a difference from the reference signal is taken to improve the SN ratio (Patent Document 4). By these signal processing methods, the SN ratio can be improved in each device.

JP-A-9-219563 WO01 / 075414 JP 2002-318190 A JP 2006-234893 A

R. Sidhu, N. Duan, JCCampbell, ALHolmes, "A Long-Wavelength Photodiode on InP Using Lattice-Matched GaInAs-GaAsSb Type-II Quantum Wells", IEEE Photonics Technology Letters, Vol. 17, No. 12 (2005 ), pp.2715-2717

  However, the dark current cannot be made zero with the above light receiving element. For this reason, even if an optical chopper is used, when the amount of light is small, the SN ratio is lowered and a clear image cannot be obtained. Furthermore, in the method of always measuring the dark signal and obtaining the difference, and the method of obtaining the difference from the reference signal, if a modulation function is added, a separate dedicated component is required, contrary to downsizing of the device, and Deteriorating economy by increasing the number of parts. Furthermore, it is necessary to optimize the optical system.

  An object of the present invention is to provide a miniaturized near-infrared image sensor capable of obtaining a signal with a high S / N ratio or a clear image even in the presence of a dark current.

  The near-infrared image sensor of the present invention is a near-infrared light image sensor having a wavelength of 1.2 μm or more. One or a plurality of light receiving elements formed on the InP substrate and having a light receiving layer for receiving near infrared light, a modulation unit positioned closer to the near infrared light incident side than the light receiving elements, and the electric power of the light receiving elements A signal processing unit for processing the signal. The band gap wavelength of the light receiving layer is 1.2 μm or more and 3 μm or less, and the modulation unit is formed of a micro electro mechanical system (MEMS) mainly composed of silicon, and includes one or more The light receiving element is covered and integrated with the light receiving element, and the signal processing unit includes a signal reading circuit that reads a signal of the light receiving element and a signal detection unit that detects a signal from the light receiving element. And

  With the above configuration, since the input light is modulated, it is possible to improve the SN ratio by removing dark current components that are not modulated even when the dark current is large. In addition, since the modulation unit is formed of silicon MEMS, it is transparent at a wavelength of 1.2 μm or more, and is integrated in a laminated structure so as to cover the light receiving element. Maintenance costs can be reduced. For the integration by the laminated structure of the modulation unit, the terminal of the modulation unit is joined to the terminal of the light receiving element and / or the terminal of the signal reading circuit (multiplexer) that is assumed to be arranged by flip chip bonding. Can be realized. The plurality of light receiving elements are usually arranged in an array. However, in the case of a one-dimensional array, a plurality of wavelengths can be detected simultaneously in combination with a spectroscopic prism or a diffraction grating. Further, if the object is moved, it can be detected two-dimensionally. In the case of a two-dimensional array, detection can be performed in two dimensions in an instant. If the object is moved in combination with the spectroscopic prism, a plurality of wavelengths can be detected two-dimensionally at the same time. Here, the fact that silicon is transparent to light having a wavelength of 1.2 μm or more means that the transmittance of light having a wavelength of 1.2 μm or more is 50% or more.

  The modulation unit can be formed by a silicon substrate and a silicon MEMS provided on the silicon substrate. Accordingly, electrodes and circuits for driving can be provided on the silicon substrate, and it becomes easy to apply modulation by MEMS while reducing the size. Further, as described above, silicon is transparent to near-infrared light having a wavelength of 1.2 μm or more, which is effective for preventing deterioration of incident light.

  The signal processing unit can read out an output generated in the light receiving element by a signal reading circuit and detect lock-in by the signal detection unit. Thus, the near-infrared image sensor can be reduced in size by integrating the signal readout circuit and the signal detection unit. A CMOS or CCD can be used for the signal readout circuit. Since the signal processing unit can be formed on a silicon substrate with one chip, it does not substantially deteriorate light having a wavelength of 1.2 μm or more, so it can be placed at any position of the near infrared image sensor. Good. Normally, in order to realize the conductive connection between the pixel electrode and the readout electrode with a simple structure, the back surface of the InP substrate is used as the incident surface, the surface side of the epitaxial layer is used as the mounting surface, and the light receiving element is disposed on the opposite side of the incident side. Is done.

  One or more light receiving elements have individual impurity regions formed by selective diffusion of impurity elements and having non-selected peripheral portions covered with a mask, and the tips of the individual impurity regions are pn junctions Can be taken. As a result, crosstalk between adjacent light receiving elements can be reduced, and the signal-to-noise ratio of the signal can be increased. In addition, the dark current can be reduced, and the cross effect of the dark current on the signal current can be reduced. The near-infrared image sensor of the present invention can obtain a signal-to-noise ratio that is equal to or higher than a predetermined level even when there is dark current. Further, when the light receiving element electrode (pixel electrode) is provided on the front side, the contact resistance between the electrode and the impurity introduction surface can be reduced.

  The light receiving layer may be formed of a multiple quantum well structure, and the light receiving element may have a diffusion concentration distribution adjusting layer for adjusting the impurity concentration distribution on the impurity introduction surface side of selective diffusion in contact with the light receiving layer. . The combination of the multiple quantum well structure and the diffusion concentration distribution adjustment layer reduces the interfacial resistance with the electrode and enables ohmic contact, while maintaining the original function of the quantum well structure without impairing the crystallinity of the quantum well structure. Therefore, it is possible to expand the stable light receiving sensitivity up to a long wavelength range.

The impurity concentration in the light receiving layer can be 5e16 cm ⁻³ or less. This makes it possible to reliably form a pn junction necessary for light reception without impairing the crystallinity of the quantum well structure.

  The band gap energy of the diffusion concentration distribution adjusting layer can be made smaller than that of InP. Thereby, the electrical resistance of the diffusion concentration distribution adjusting layer can be made relatively small.

  The multiple quantum well structure is a type II multiple quantum well structure in which (InGaAs / GaAsSb) is paired, the impurity element is zinc (Zn), and the diffusion concentration distribution adjusting layer can be formed of InGaAs. As a result, a light receiving layer that can reliably receive light up to the upper limit of 3 μm on the long wavelength side can be obtained by the existing MBE method or the like. Further, a p-type region can be formed using a p-type impurity Zn which is easy to handle and a pn junction can be formed at the tip. Further, a pixel electrode having a small contact resistance can be provided on the introduction surface of the selective diffusion of Zn. Furthermore, by forming the diffusion concentration distribution adjusting layer of InGaAs, it is possible to form a layer having a low electric resistance while lattice matching with the multiple quantum well of the light receiving layer.

Lattice matching of the InP substrate, the light receiving layer, and the diffusion concentration distribution adjusting layer is defined as a _i and the difference between the lattice constant of the InP substrate, the light receiving layer, and the diffusion concentration distribution adjusting layer is Δa _ij. The degree (Δa _ij / a _i ) can be 0.002 or less. Here, i or j = 1 to 3 is assigned to the InP substrate, the light receiving layer, and the diffusion concentration distribution adjusting layer. When the light receiving layer has a multiple quantum well structure, the lattice matching degree is satisfied for each of the pairs. By satisfying the above condition of lattice matching, the lattice defect density can be reduced and the flatness at the interface of each layer can be improved. By obtaining this excellent crystallinity, dark current can be reduced.

  The near-infrared image sensor described above can be used as a food inspection device, a medical camera, or a monitoring device. Thereby, useful and highly sensitive information can be obtained about the biogenic substance having an absorption spectrum peak at a wavelength of 1 μm to 3 μm. This information makes it possible to detect the freshness and quality of food, the presence or absence of lesions for medical use, and abnormalities in components. In addition, with regard to the monitoring device, it is possible to detect a human intrusion with space light without a light source at night.

  According to the near-infrared image sensor of the present invention, a signal with a high S / N ratio or a clear image can be obtained even in the presence of a dark current, and the size can be reduced.

It is sectional drawing which shows the near-infrared image sensor in Embodiment 1 of this invention. It is a figure which shows the modulation part in the near-infrared image sensor of FIG. It is a figure for demonstrating the lock-in detection in the near-infrared image sensor of FIG. The modification of the near-infrared image sensor of Embodiment 1 of this invention is shown, (a) is a perspective view of the modulation | alteration part different from FIG. 2, (b) is sectional drawing which follows the IVB-IVB line | wire of (a). is there. It is sectional drawing which shows the near-infrared image sensor in Embodiment 2 of this invention. It is a figure which shows the density distribution of the p-type impurity in the light receiving element of FIG. It is a top view of the near-infrared image sensor (imaging device) of FIG. It is a figure for demonstrating the signal processing of CMOS. It is a figure explaining the circuit which carries out voltage conversion and amplifies the electric charge stored from the photodiode. It is a figure for demonstrating the signal processing of the digital camera incorporating the near-infrared image sensor of FIG. It is a figure which shows the biological component detection apparatus using the near-infrared image sensor in Embodiment 3 of this invention.

(Embodiment 1)
FIG. 1 is a diagram showing a near-infrared image sensor 100 according to Embodiment 1 of the present invention. The near-infrared image sensor 100 includes a light receiving element 10, a modulation unit 50 positioned on the incident side of the light receiving element 10, and a signal processing unit 70 including a signal readout circuit positioned on the opposite side to the incident side of the light receiving element 10. Is done.

(Light receiving element)
A light receiving layer 3 having a band gap wavelength of 1.8 μm to 3 μm is formed on the InP substrate 1, and a first conductivity type region 6 is formed for each pixel P by selective diffusion of the first conductivity type impurities. A pn junction, pi junction, or ni junction is formed at the tip interface 15 on the light receiving layer 3 side of the first conductivity type region 6 included in the pixel P. A reverse bias voltage is applied to the junction, a depletion layer is formed on the light receiving layer 3 side, and incident light (signal light) is photoelectrically converted in the depletion layer. The first conductivity type region 6 is formed by selectively diffusing a first conductivity type impurity from the surface of the window layer 5. The first conductivity type electrode 11 constituting the pixel electrode is formed so as to be in ohmic contact with the window layer side surface of the first conductivity type region 6. Further, any part of the InP substrate 1 or the light receiving layer 3 or the buffer layer 2 is of the second conductivity type, and the second conductivity type region is set to the ground potential, and is in ohmic contact with a second conductivity side electrode (not shown). Is done. Since the InP substrate 1 substantially transmits light in the wavelength range to be received, it can be mounted on the epitaxial surface side and the back surface of the InP substrate 1 can be used as the incident surface.

(Signal processing part)
A signal processing unit 70 including a signal readout circuit is arranged below the light receiving element 10 formed on the InP substrate 1 (on the side opposite to the input light incident surface). The first conductive side electrode 11 is conductively connected for each pixel to the readout electrode 71 of the signal readout circuit, and the common ground electrode or the second conductive side electrode of the light receiving element 10 is conductively connected to the ground electrode (not shown). Is done. For the signal readout circuit, it is preferable to use a complementary metal oxide semiconductor (CMOS), a charge coupled device (CCD), or the like. In the pixel P, light is received for a predetermined time to generate photoelectric charge, and this photoelectric charge is accumulated in each pixel. The accumulated photoelectric charge is read as a current by the signal readout circuit. Depending on the type of the signal readout circuit, it can be processed as it is (CCD), or converted into a voltage and amplified at each pixel (CMOS). Until the predetermined stage, the type of electric signal is made different. In either case, a voltage is output from the signal readout circuit. A processing method of output from the pixel in the CMOS will be described in detail in Embodiment 2. The conductive connection between the light receiving element 10 and the electrodes of the signal readout circuit is not shown in FIG. 1, but it is preferable to use In bumps, an anisotropic conductive film, or the like.

(Modulation section)
The modulation unit 50 may be anything as long as it is composed of MEMS mainly composed of silicon and can modulate incident light at a controlled constant frequency. FIG. 2 shows a cantilever type modulation device formed by MEMS. The modulation device 50 is formed on a silicon substrate 51, and a movable portion 55 or a beam made of silicon is fixedly supported at one end by a support portion 52. The beam 55 vibrates by receiving a force periodically by an oscillating voltage or an oscillating electric field from the lower electrode 53. The gap d between the silicon substrate 51 and the beam 55 is, for example, ½ wavelength or one wavelength. When the beam does not vibrate, the incident light is partially reflected on the surface of the silicon substrate, and resonance occurs to form a standing wave in the gap d. Therefore, the incident light reaching the transmitted light or the light receiving element 10 is Less. When a vibration voltage is applied to the lower electrode 53 to vibrate the beam 55, the resonance condition in the gap d is broken and the amount of transmitted light increases.
As shown in FIG. 1, the modulation unit 50 is fixed to the light receiving element by a fixing unit 19 such as a solder bump or an epoxy adhesive. Although not shown, control of the oscillating voltage applied to the lower electrode 53 is performed by a drive circuit formed on the silicon substrate 51, and information on modulation drive is transmitted to the lock-in detection unit of the signal processing unit 70.

  FIG. 3 shows the illuminance due to the incident light, the current (charge) caused by the dark current, and the accumulation of the photo charge due to the modulated light and the dark current, the voltage conversion, And it is a schematic diagram which shows the synthesized voltage signal after amplification. Referring to FIG. 1, the combined voltage signal is sent to the lock-in detection unit of signal processing unit 70 to detect lock-in. The dark current is caused by the photoelectric conversion portion of the epitaxial layered body of the InP substrate 1, the light receiving layer 3, and the pixel P. Therefore, the dark current is not modulated by the modulation portion 50, and there is a ripple (pulsation), but a substantially constant current (charge) ) That is, a voltage is generated. Noise that is not limited to dark current and is not modulated is removed as a DC component by lock-in detection. For this reason, it is possible to select a signal caused by incident light with high accuracy and obtain a signal having a high S / N ratio.

  In the lock-in detection unit, it is easy to remove the DC component from the synthesized signal by frequency analysis, so if the input optical signal and dark current or noise interact and are included in the same frequency The frequency component caused by noise can be removed by analyzing the frequency as small. For example, in the case of a near-infrared image sensor with one light receiving element, the frequency of the readout cycle can be about 200 kHz. By setting the modulation frequency of the modulated light for lock-in detection to the same frequency, it is possible to read only the signal, excluding the noise component, by reading the DC component of the synthesized signal. Also, when the dark current (noise) increases proportionally according to the intensity of the input light, the modulation signal can have an arbitrary waveform, and the pattern is known, so that many noise components are generated from the synthesized signal. Can be removed. As a result, by analyzing the frequency of the synthesized signal in lock-in detection, a signal with a high S / N ratio due to incident light can be obtained.

The circuit configuration may be such that the signal processing unit 70 including the signal readout circuit shown in FIG. 1 includes a lock-in detection unit. The signal readout circuit is usually formed on a silicon substrate, and it is easy to build a lock-in detection unit therein. As a result, it becomes easy to reduce the size of the near-infrared image sensor 100.
Further, as described above, both the silicon substrate 51 and the silicon beam 55 have a transmittance of 50% or more for light having a wavelength of 1.2 μm or more, so that the modulation is performed without significantly reducing the signal intensity level. Can do. Near-infrared light does not pass through wiring, lower electrodes, solder bumps, etc. made of metal even though silicon is transmitted. However, the modulation unit 50 is not entirely covered with the metal portion, and only affects a part of the modulation unit 50. Therefore, the above-described “modulated incident light → synthetic voltage including dark current”. There is no change in the scheme of “signal formation → lock-in detection”.

(Modification of Embodiment 1)
FIG. 4 is a diagram illustrating a modification example in which a MEMS comb shutter is used instead of the modulation device using the MEMS cantilever 55 of the modulation unit 50 in the first embodiment. (A) is a perspective view, (b) is sectional drawing which follows the IVB-IVB line. In this modulation section 50, one lower electrode 53 of the comb teeth is provided on the silicon substrate 51, and the upper movable section 55 or the other comb teeth 55 is formed of polysilicon, and an aluminum electrode is deposited thereon. Then, it is supported by the support column 57 and floats. When an oscillating voltage is applied to the lower electrode 53, the upper comb teeth 55 receive a force in the horizontal direction and vibrate horizontally. Thus, a shutter that opens and closes in a vibrating manner can be formed, and on-off modulation can be applied to incident light.

(Embodiment 2)
FIG. 5 is a diagram showing a part of the near-infrared image sensor 100 according to Embodiment 2 of the present invention. The target wavelength range is the near-infrared range, the InP substrate 1 is used, and the light receiving layer 3 has a multiple quantum well structure of InGaAs and GaAsSb that has a band gap wavelength including subbands of 1.8 μm to 3 μm. Is used. A group III-V semiconductor laminated structure (epitaxial wafer) having the following configuration is provided on the InP substrate 1.
(InP substrate 1 / InP buffer layer 2 / light-receiving layer 3 having a multiple quantum well structure of InGaAs and GaAsSb / InGaAs diffusion concentration distribution adjusting layer 4 / InP window layer 5)
In the p-type region 6 located so as to reach the light-receiving layer 3 having the multiple quantum well structure from the InP window layer 5, the p-type impurity Zn is selectively diffused from the opening of the selective diffusion mask pattern 36 of the SiN film. Formed with. By diffusing p-type impurities using the selective diffusion mask pattern 36 of the SiN film, the p-type region 6 and, as a result, a pn junction can be formed inside the peripheral edge for each pixel P. In the selective diffusion, the p-type impurity is diffused and introduced into the periphery of the non-selected peripheral portion covered with the mask in a planar manner so that the pn junction is exposed at the end face of the light receiving element in the case of one light receiving element. In the case of a plurality of light receiving elements, they are isolated from adjacent light receiving elements. As a result, leakage of photocurrent and crosstalk are suppressed.

  A p-side electrode 11 made of AuZn is provided in the p-type region 6, and an n-side electrode 12 made of AuGeNi is provided in ohmic contact with the InP buffer layer 2. The n-side electrode 12 is common to all the pixels, is a ground electrode, and is conductively connected to the ground electrode 72 of the CMOS 70a by an In bump (not shown). The p-side electrode 11 constituting the pixel electrode is conductively connected to the readout electrode 71 of the CMOS 70a. On the back surface (incident surface) of the InP substrate 1, an antireflection film 35 of SiON is also provided. The antireflection film 35 may be omitted.

In the light-receiving layer 3 having the multiple quantum well structure, a pn junction (or pi junction) 15 is formed at a position corresponding to the boundary front of the p-type region 6, and between the p-side electrode 11 and the n-side electrode 12. By applying a reverse bias voltage, a depletion layer is generated more widely on the side where the n-type impurity concentration is low (n-type impurity background). The background in the light-receiving layer 3 having the multiple quantum well structure is about 5 × 10 ¹⁵ / cm ³ or less in terms of n-type impurity concentration (carrier concentration). The position 15 of the pn junction is determined by the intersection of the background (n-type carrier concentration) of the light-receiving layer 3 of the multiple quantum well and the concentration profile of the p-type impurity Zn. That is, the position shown in FIG. In the diffusion concentration distribution adjusting layer 4, the concentration of the p-type impurity selectively diffused from the surface 5 a of the InP window layer 5 sharply decreases from the high concentration region on the InP window layer side to the light receiving layer side. Therefore, in the light receiving layer 3, an impurity concentration of 5 × 10 ¹⁶ / cm ³ or less can be easily realized.

  Since the light receiving element 10 targeted by the present invention seeks to have light receiving sensitivity from the near infrared region to the long wavelength side, a material having a band gap energy larger than the band gap energy of the light receiving layer 3 is used for the window layer. It is preferable to use it. For this reason, InP, which is a material having a band gap energy larger than that of the light receiving layer and having a good lattice matching, is usually used for the window layer. InAlAs having substantially the same band gap energy as InP may be used.

In the multi-quantum well structure, when impurities are introduced at a high concentration by selective diffusion, the crystal structure is destroyed. Therefore, it is necessary to suppress the introduction of impurities by selective diffusion. Usually, the concentration of the p-type impurity to be diffused and introduced is required to be 5 × 10 ¹⁶ / cm ³ or less. The reason why the impurity concentration of the light receiving layer is 5 × 10 ¹⁶ / cm ³ or less will be described in detail below. When the Zn concentration in the light receiving layer 3 exceeds 1 × 10 ¹⁷ cm ^{−3 because} the depth of selective diffusion of the p-type impurity (Zn) is increased, the InGaAs that constitutes the quantum well layer in the higher concentration portion. And GaAsSb atoms enter each other and are disturbed, destroying the superlattice structure. The crystal quality of the destroyed portion is lowered, and the device characteristics are deteriorated, such as an increase in dark current. Here, the Zn concentration is usually measured by SIMS analysis (Secondary Ion Mass Spectroscopy), but it is difficult to analyze the concentration of 10 ¹⁷ cm ⁻³ or 10 ¹⁶ cm ^−3. Large measurement error occurs. The detailed description above is a discussion of the Zn concentration with double or half accuracy, but it comes from this measurement accuracy. Therefore, for example, it is difficult to discuss the difference between 5 × 10 ¹⁶ / cm ³ and 6 × 10 ¹⁶ / cm ^{3 in} terms of measurement accuracy, and there is no great meaning. In order to stably obtain the low p-type impurity concentration with good reproducibility in actual production, a diffusion concentration distribution adjusting layer 4 made of InGaAs is provided on the light receiving layer 3. In the diffusion concentration distribution adjusting layer 4, when the thickness range on the light receiving layer side becomes a low impurity concentration as described above, the electrical conductivity in the low impurity concentration range decreases or the electrical resistance increases. When the electrical conductivity in the low impurity concentration range in the diffusion concentration distribution adjusting layer 4 is lowered, the responsiveness is lowered and, for example, a good moving image cannot be obtained. However, when the diffusion concentration distribution adjusting layer is formed of a material having a band gap energy smaller than the band gap energy equivalent to InP, specifically, a group III-V semiconductor material having a band gap energy of less than 1.34 eV, Even at low concentrations, the electrical conductivity is not significantly reduced. As a group III-V semiconductor material satisfying the requirements of the diffusion concentration distribution adjusting layer, for example, InGaAs can be cited.
When a material having a narrow band gap energy is used for the diffusion concentration distribution adjusting layer, an increase in electrical resistance can be suppressed even if the impurity concentration is low. Since the response speed to reverse bias voltage application or the like is considered to be determined by the CR time constant due to the capacitance and electric resistance, the response speed can be shortened by suppressing the increase in electric resistance R as described above.

  In the present embodiment, the multiple quantum well structure is type II. In the type I quantum well structure, in the case of a light receiving element that has a light receiving sensitivity in the near infrared region while sandwiching a semiconductor layer having a small band gap energy between semiconductor layers having a large band gap energy, a semiconductor layer having a small band gap energy is used. The upper limit wavelength (cutoff wavelength) of the light receiving sensitivity is determined by the band gap. That is, transition of electrons or holes due to light is performed in a semiconductor layer having a small band gap energy (direct transition). In this case, the material for extending the cutoff wavelength to a longer wavelength region is very limited in the III-V compound semiconductor. On the other hand, in the type II quantum well structure, when two different semiconductor layers having the same Fermi energy are alternately stacked, the conduction band of the first semiconductor and the valence band of the second semiconductor are obtained. The upper limit of the wavelength (cutoff wavelength) of the light receiving sensitivity is determined. That is, transition of electrons or holes by light is performed between the valence band of the second semiconductor and the conduction band of the first semiconductor (indirect transition). For this reason, the energy of the valence band of the second semiconductor is made higher than that of the first semiconductor, and the energy of the conduction band of the first semiconductor is made lower than the energy of the conduction band of the second semiconductor. By doing so, it is easier to realize a longer wavelength of light receiving sensitivity than in the case of direct transition in one semiconductor.

  Next, a method for manufacturing the light receiving element 10 shown in FIG. 5 will be described. An InP buffer layer 2 or InGaAs buffer layer 2 having a thickness of 2 μm is formed on the n-type InP substrate 1. Next, a light-receiving layer 3 having an (InGaAs / GaAsSb) multiple quantum well structure is formed. The thickness of the InGaAs layer and the GaAsSb layer forming the unit quantum well structure is 5 nm, respectively, and the number of pairs (the number of repetitions of the unit quantum well) is 300. Next, an InGaAs layer having a thickness of 1 μm is epitaxially grown on the light receiving layer 3 as a diffusion concentration distribution adjusting layer 4 at the time of introducing Zn diffusion, and finally, an InP window layer 5 having a thickness of 1 μm is epitaxially grown. Both the light receiving layer 3 and the diffusion concentration distribution adjusting layer 4 are preferably epitaxially grown by MBE (Molecular Beam Epitaxy) method. The InP window layer 5 may be epitaxially grown by the MBE method, or after the diffusion concentration adjusting layer 4 is grown, the InP window layer 5 may be taken out from the MBE apparatus and epitaxially grown by the MOVPE (Metal Organic Vapor Phase Epitaxy) method. . Further, all epitaxial growth may be performed by the MOVPE method.

The InP buffer layer 2 or the InGaAs buffer layer 2 may be non-doped, or may be doped with an n-type dopant such as Si at about 1 × 10 ¹⁷ / cm ³ . The InGaAs / GaAsSb multiple quantum well structure light-receiving layer 3, InGaAs diffusion concentration distribution adjusting layer 4, and InP window layer 5 are preferably non-doped, but a very small amount of n-type dopant such as Si (for example, 2 × 10 ¹⁵ / cm 2). ^{About 3} ) Doping may be performed. Further, a high-concentration n-side electrode forming layer for forming an n-side electrode doped with n-type dopant by about 1 × 10 ¹⁸ / cm ³ may be inserted between the InP substrate 1 and the buffer layer 2. . The InP substrate 1 may be a Fe-doped semi-insulating InP substrate. In this case, an n-side electrode forming layer doped with n-type dopant of about 1 × 10 ¹⁸ / cm ³ is inserted between the semi-insulating InP substrate 1 and the buffer layer 2. The buffer layer 2 may be an n-side electrode formation layer.

Next, using the SiN selective diffusion mask pattern 36 formed on the surface 5a of the InP window layer 5, Zn is selectively diffused from the opening (InGaAs / GaAsSb) so as to reach the light receiving layer 3 having the multiple quantum well structure. A p-type region 6 is formed. The front tip of the p-type region 6 forms a pn junction 15. At this time, a high concentration region having a Zn concentration of about 1 × 10 ¹⁸ / cm ³ or more is limited to the InGaAs diffusion concentration distribution adjusting layer 4. That is, the high-concentration impurity distribution continues from the surface 5a of the InP window layer 5 in the depth direction to the InGaAs diffusion concentration distribution adjustment layer 4, and further 5 × at a deeper position in the diffusion concentration distribution adjustment layer 4. It decreases to 10 ¹⁶ / cm ³ or less (see FIG. 6). The Zn concentration distribution in the vicinity of the pn junction 15 is a distribution indicating an inclined junction.

  The one-dimensional or two-dimensional array of the light-receiving elements 10, that is, the light-receiving element array shown in FIG. 5 is planarly selected so that Zn is selectively diffused without performing mesa etching for element separation (inside the peripheral edge of the light-receiving element). The adjacent light receiving elements are separated from each other by diffusion limited to the periphery. That is, the Zn selective diffusion region 6 is a main part of one light receiving element 10 and forms one pixel P, but a region where Zn is not diffused separates each pixel. For this reason, it is possible to suppress dark current without being damaged by crystals accompanying the mesa etching.

  FIG. 7 is a plan view of the light receiving element 10 of the near infrared image sensor 100 of FIG. 5 as viewed from the light incident side. FIG. 5 is a cross-sectional view (including the modulation unit 50 and the CMOS 70a) taken along the line VV in FIG. In FIG. 5, the SiN selective diffusion mask pattern 36 is left as it is together with the protective SiON film 37 formed thereon. As shown in FIG. 7, the pixels P or sensor units are provided at a pitch of 25 μm in length and width of 20 mm and 16 mm, and a total of 640 × 512 = 327,680 is arranged. Needless to say, this pixel arrangement is merely an example and can be changed according to product specifications.

  Both the p-part electrode 11 electrically connected to each p-type region 6 of the epitaxial layer and the n-part electrode 12 provided in the n-type buffer layer 2 located directly on the common InP substrate 1 are used for signal processing. It is connected to the CMOS 70a included in the part by a bonding bump 33 such as a solder bump. The p-part electrode 11 is made of AuZn, and the n-part electrode 12 is made of AuGeNi so as to ensure ohmic contact. In addition to the above AuZn-based metal, a TiPt-based metal may be used for the p-part electrode. In the above structure, the pixels P or photodiodes arranged in a matrix form a photocurrent generating portion of the light receiving portion of each light receiving element.

  The modulation unit 50 may be a resonance modulation device using the MEMS cantilever shown in FIG. 2 according to the first embodiment, or may be a modulation device using the MEMS comb shutter shown in FIG. Since both modulators are microelectromechanical systems mainly composed of silicon provided on a silicon substrate, they are transparent to near infrared light having a wavelength of 1 μm or more and have a high response speed. Further, since the MEMS drive control unit can be easily provided on the silicon substrate 51, the near infrared image sensor 100 can be obtained by fixing the modulation unit 70 to the light receiving element 10 with a solder bump or an epoxy adhesive. Can be miniaturized.

  Incident light is introduced through the AR film 35 formed on the back surface of the InP substrate 1 and received in a depletion layer protruding from a pi junction or a pn junction which is an interface between the p-type region 6 and the light receiving layer 3. A reverse bias voltage for forming the depletion layer is applied between the n-side electrode 12 and the p-side electrode 11. The depletion layer acts as a capacitor and accumulates photoelectrically converted charges. Of the hole-electron pairs generated in the light-receiving portion by photoelectric conversion, holes are read from the p-type region 6 as signal charges to the read electrode 71 of the CMOS 70 via the p-side electrode 11 and all of the holes are scanned by pulse drive scanning. The above readout is sequentially performed on the pixel portion.

  Next, lock-in detection of the near-infrared image sensor will be described. In the case of a moving image, since 30 frames are scanned per second, it takes 33 ms per frame. For example, when applied to an imaging device of 80,000 pixels of 320 × 256 pixels (pixels), the time per pixel is 0.4 μs, and reading is performed at a frequency of 2.5 MHz. In lock-in detection, reference light (modulated light) is irradiated at a modulation frequency of 2.5 MHz to generate combined light. By applying a narrow band low-pass filter to the combined light and extracting the DC component, noise can be removed and only the signal light can be read. Since the clock of the lock-in reference signal is from a normal level of 200 kHz to about 5 MHz for special use, the above lock-in detection can be easily realized. As another example, a case of an imaging device having 64 × 64 pixels will be described. In this imaging apparatus, the readout frequency is 130 kHz. Therefore, lock-in detection can be performed using a normal level lock-in device. In any of the above examples, as a result of the lock-in detection, the S / N ratio of the signal can be improved for each light receiving element. Further, when synchronizing the frame period and the reference signal, it is preferable to set the modulation frequency to 30 Hz to 200 Hz, depending on the degree of request for high-speed processing of image analysis.

FIG. 8 is a diagram for explaining a signal processing method in a general CMOS 70a. FIG. 9 is an enlarged view of the A portion including the switch portions S ₁ and S ₂ in FIG. 8, and is a diagram for explaining a mechanism for converting the amount of holes generated in each pixel (photodiode) P into a voltage and amplifying it. is there. In FIG. 8, a horizontal line is first selected from among the photodiodes P of the pixel portion arranged in a matrix by a switching signal from a Y address circuit that selects a horizontal line, and then, the X address circuit includes the horizontal line. One photodiode is selected. Then, the charge accumulated for a predetermined time as described above is read from the one photodiode P to the output signal line. The photodiodes P are selected one by one according to the scanning order.

  The CMOS converts the signal charge from the photodiode P into a signal voltage for each photodiode P, amplifies it, and sequentially reads out the signal voltage to obtain an electrical signal. This includes dark current. As shown in FIG. 3, this signal voltage is obtained by voltage-converting and amplifying the sum of the photocharge (modulated photocharge) by the modulated light obtained by modulating the incident light and the charge of the dark current. It is. That is, it is a composite voltage signal that has been modulated including dark current. In a normal CMOS, electrons out of electron-hole pairs generated by photoelectric conversion are sent to a current-voltage conversion unit. However, in the imaging device of FIG. An example of the configuration after such a change is shown in FIG. In the CMOS, when the voltage is amplified in each pixel portion, the black level varies depending on the variation of the threshold level of the MOS transistor, resulting in noise. A voltage corresponding to a different black level is clamped by a CDS (Corelated Double Sampling) circuit in FIG. 8 to obtain a difference from the signal level, thereby eliminating the influence of the difference in the threshold level of the MOS transistor. .

  The characteristics when a CMOS is used for the signal processing unit are as follows. (1) Low power consumption can be realized. (2) Random access is possible. That is, not only can the signals be sequentially read out from the photodiodes 10p of each pixel portion, but also the signals from the respective photodiodes can be read out at random. (3) It can be operated with a single power source and a single clock. For this reason, the number of wirings between the epitaxial laminated body 10 in which the photodiode P is formed and the CMOS 70 is minimized, which is advantageous in terms of radiation noise. (4) The entire control circuit unit including a signal processing unit such as a drive unit for scanning and reading out signal charges can be configured on a single chip.

  FIG. 10 is a diagram showing a signal processing process of a digital camera incorporating the near-infrared image sensor 100 of the present embodiment. In FIG. 10, the near-infrared image sensor 100 of the present embodiment includes a modulation unit 50 that modulates near-infrared light incident on the light-receiving element 10, the light-receiving element 10 including the light-receiving layer 3, and the light-receiving element 10. And a signal processing unit 70 including a CMOS that reads out a signal from each pixel (photodiode) P and processes the signal. Since the signal processing unit 70 is formed on a semiconductor substrate such as silicon, process circuits subsequent to the signal readout circuit, such as a detection unit that performs CDS / AGC (Automatic Gain Control) and lock-in detection in FIG. Can be easily formed. The silicon substrate on which the signal processing unit 70 including the CMOS is formed is inexpensive, and the processing using the silicon substrate has an abundant track record and the processing equipment is sufficient, so it is suitable for mass production economically. ing.

  In FIG. 10, a signal from which predetermined noise has been removed by the CDS circuit is input to the AGC circuit. In the AGC circuit, the gain of the amplifier is controlled by a microcomputer (CPU: Central Processing Unit), and the intensity level of the signal necessary for the A / D converter is obtained in conjunction with the aperture driving mechanism. The detection unit may be disposed in front of the AGC circuit. The detection unit performs lock-in detection, and when the noise caused by dark current that cannot be handled by a CDS circuit or the like is large and the input optical signal is small, a modulation signal is applied to generate a composite signal with dark current that is not modulated. After formation, a signal with a high S / N ratio can be obtained. In other words, when modulation is not performed, when dark current is large, an electric signal is weak, and it is buried in noise (low S / N ratio), modulation is performed and frequency analysis is performed by lock-in detection. It is possible to detect a signal caused by the N ratio input light.

In the near-infrared image sensor 100 shown in FIG. 10, as described above, the CDS / AGC circuit, the detection unit, the A / D converter, and the like are formed together on the silicon substrate on which the signal processing unit 70 including the CMOS is formed. Can be done. It can be considered that the CMOS includes the circuit section including the detection section.
The modulation unit 50 modulates near-infrared light incident on the photodiode array 10 formed on the InP substrate 1 with the silicon substrate 51 interposed. Silicon has a small absorption rate of light in the near infrared region, and transmits near infrared light. Since the MEMS including the silicon substrate 1 is thin, the near-infrared image sensor 100 including the modulation unit 50 / the light receiving element 10 / the signal processing unit 70 can be realized with a compact configuration. Further, by forming a control circuit on the silicon substrate on which the signal processing unit 70 is formed in accordance with the measuring instrument using the near-infrared image sensor 100, the measuring device can be downsized, and the control circuit can be adapted to the measuring device. If it is made flexible, it can be used in the measuring device with little or no modification.

(Modification of Embodiment 2)
The photodiode P shown in FIG. 5 exemplifies a conductive structure that reads holes. That is, it has a conductive structure in which holes are read from the p-type region 6 to the read electrode 71 of the CMOS 70. However, the conductivity type of each part in FIG. 5 can be reversed to have a conductivity type structure in which electrons are read out to the readout electrode 71 of the CMOS 70. In the embodiment of the present invention, a device using a CMOS for the signal processing unit is illustrated, but a CCD may be used instead of the CMOS.

(Embodiment 3)
FIG. 11 is a diagram showing a near-infrared image sensor 100 according to Embodiment 3 of the present invention. The near-infrared image sensor 100 is used by being incorporated in a biological component detection device. A characteristic is that a living body insertion groove 67a is provided in a part of the housing 67, and a blood glucose level is detected using transmitted light of the living body inserted into the living body insertion groove 67a. The living body part to be inserted is assumed to be the tip of the shoulder, for example, an arm or a palm, and the living body insertion groove 67a having the maximum size among them can be used. The living body insertion groove 67a specializing in the earlobe may be used.

The path of near infrared light is as follows.
Light source 63-> condensing lens 87-> reflecting mirror 66-> condensing lens 87-> irradiation optical fiber 81-> detection site-> light receiving end 82a-> press adjusting actuator 82b-> information-equipped optical fiber 82-> condensing lens 87-> diffraction grating 91 → Near-infrared image sensor 100 → Display unit 85 (see FIG. 11)
The position of the pixel of the light receiving element array 10 in the near infrared image sensor 100 corresponds to the wavelength of light diffracted by the diffraction grating 91 according to the wavelength. That is, the continuous light-receiving element array receives the diffracted light of continuous wavelengths at a predetermined wavelength pitch. A spectral prism can be used instead of the diffraction grating 91. As such a light receiving element array, a one-dimensional array of light receiving elements arranged along the direction of diffraction according to the wavelength is sufficient, but a two-dimensional array may be used. In the case of a two-dimensional array, the apparatus configuration is easy because alignment does not have to be precise along one direction.

In FIG. 11, the lower part of the little finger when the hand sword is made with a palm can transmit light without interposing a bone, which is effective in measuring the blood sugar level. The living body insertion groove 67a is not particularly required to be focused on the palm portion below the little finger, and can be aligned by the pressure adjusting actuator 82b or the like. As a result, the patient can easily and accurately measure the blood glucose level.
The light-receiving element array in the near-infrared image sensor 100 can receive light up to a long wavelength region in the near-infrared region, and can improve measurement accuracy.
As the light source, a halogen lamp or the like is preferably used. However, in the case of this biological component detection device, it is preferable to use a continuum light source or an LED that generates little heat.

  Although the embodiments and examples of the present invention have been described above, the embodiments and examples of the present invention disclosed above are merely examples, and the scope of the present invention is the implementation of these inventions. It is not limited to the form. The scope of the present invention is indicated by the description of the scope of claims, and further includes meanings equivalent to the description of the scope of claims and all modifications within the scope.

  According to the near infrared image sensor of the present invention, when the noise level derived from the photoelectric conversion unit is high or the dark current is high, the modulation unit is simple and compact and transparent to near infrared light having a wavelength of 1 μm or more. By modulating the incident light, a signal having a high S / N ratio can be detected. For this reason, a clear image and high sensitivity can be detected with a compact device.

1 InP substrate, 2 n-type buffer layer, 3 light-receiving layer, 4 diffusion concentration distribution adjusting layer, 5 window layer, 5a surface of window layer, 6 p-type region (first conductivity type region), 10 light-receiving element (light-receiving element array) ), 11 p-side electrode, 12 n-side electrode (common ground electrode), 15 pn junction, 19 fixing part (bump, adhesive), 35 AR film, 36 selective diffusion mask pattern, 37 SiON film, 50 modulation part, 51 silicon substrate, 52 support part, 53 lower electrode, 55 movable part (beam, comb tooth), 57 support column, 63 light source, 66 reflector, 70 signal processing part, 70a CMOS (signal readout circuit), 71 readout electrode, 72 ground electrode, 81 irradiating optical fiber, 82 information mounting optical fiber, 82a light receiving end, 82b pressure adjusting actuator, 85 display unit, 87 condenser lens, 91 diffraction grating, 100 near infrared light Jisensa, P pixel (photodiode), S _1, S ₂ switch.

Claims (10)

A near infrared light image sensor having a wavelength of 1.2 μm or more,
One or a plurality of light receiving elements formed on an InP substrate and having a light receiving layer for receiving the near infrared light;
A modulator located on the incident side of the near infrared light from the light receiving element;
A signal processing unit for processing an electrical signal of the light receiving element,
The band gap wavelength of the light receiving layer is 1.2 μm or more and 3 μm or less,
The modulator is formed of a micro electro mechanical system (MEMS) mainly composed of silicon, covers the one or more light receiving elements, and is integrated with the light receiving elements.
The near-infrared image sensor, wherein the signal processing unit includes a signal reading circuit that reads a signal from the light receiving element, and a signal detection unit that detects a signal from the light receiving element.   The near-infrared image sensor according to claim 1, wherein the modulation unit is formed of a silicon substrate and a silicon MEMS provided on the silicon substrate.   The near-infrared image sensor according to claim 1, wherein the signal processing unit reads an output generated in the light receiving element by the signal reading circuit and detects lock-in by the signal detection unit. . Light receiving element.   The one or more light receiving elements have individual impurity regions formed by selective diffusion of impurity elements and having non-selected peripheral portions covered with a mask, and the tips of the individual impurity regions are pn junctions The near-infrared image sensor of any one of Claims 1-3 characterized by the above-mentioned.   The light receiving layer is formed of a multiple quantum well structure, and the light receiving element has a diffusion concentration distribution adjusting layer for adjusting the impurity concentration distribution on the impurity introduction surface side of the selective diffusion in contact with the light receiving layer. The near-infrared image sensor of Claim 4 characterized by these. The near-infrared image sensor according to claim 5, wherein an impurity concentration in the light receiving layer is 5e16 cm ⁻³ or less.   The near-infrared image sensor according to claim 5 or 6, wherein a band gap energy of the diffusion concentration distribution adjusting layer is smaller than InP.   The multiple quantum well structure is a type II multiple quantum well structure in which (InGaAs / GaAsSb) is paired, the impurity element is zinc (Zn), and the diffusion concentration distribution adjusting layer is formed of InGaAs. The near-infrared image sensor of any one of Claims 5-7 characterized by these.   The lattice constant of the InP substrate, the light receiving layer, and the diffusion concentration distribution adjusting layer is ai, and the difference between the lattice constant of the InP substrate, the light receiving layer, and the diffusion concentration distribution adjusting layer is Δaij. The near infrared image sensor according to claim 5, wherein (Δaij / ai) is 0.002 or less. The near-infrared image sensor according to claim 1, wherein the near-infrared image sensor is used as a food inspection device, a biological inspection device, a medical camera, or a monitoring device.

Images