FR3150902A1 - Interconnection structure between microcircuits - Google Patents

Abstract

According to the invention, a set of two electronic microcircuits assembled together by producing a set of electrical couplings between homologous connection points of the two microcircuits is proposed, a first microcircuit being a transducer microcircuit with at least one photoelectric transduction element by charge accumulation, and the second microcircuit being a reading microcircuit with at least one element for reading the electrical charges generated by a respective photoelectric transduction element, and the connection points comprising facing surface elements separated by a dielectric, the coupling between the two circuits being a capacitive coupling. Application in particular to the hybridization between two matrix circuits, one for transduction and one for reading, produced in different technologies. Fig. 3

Description

Interconnection structure between microcircuits

The present invention relates in general to hybrid circuits, and in particular to a new approach to electrical coupling between two circuits made in different technologies and/or requiring assembly by superposition after their manufacture.

To take a practical example, a matrix sensor is composed of pixels containing essentially a transduction element and a local readout circuit. The transduction element and the local readout circuit can be implemented either on the same substrate or on two different substrates.

The fabrication of transduction elements and readout circuits on two different substrates is typically dictated by compatibility and optimization considerations. Currently, readout circuits are fabricated using CMOS technology on a silicon substrate, while transduction elements can be fabricated using a much wider variety of techniques.

Transduction elements can, for example, be fabricated on silicon substrates using a highly optimized transduction manufacturing process. Alternatively, they can be made with less common semiconductor materials, such as III-V family semiconductors for wavelengths in the infrared range.

We will mention here in particular InGaAs substrates for applications in the near-infrared in particular for SWIR sensors (for "Short Wave Infrared" in English), InSb substrates for sensors in the mid-infrared (MWIR for "Middle Wave Infrared", and middle wave infrared), cadmium mercury telluride (MCT) for LWIR sensor applications ("Long Wave Infrared" in English, or CdTe for X-ray detection.

Such a sensor made on two assembled substrates is commonly called a "hybrid sensor", the term "hybridization" referring to the assembly process between the two substrates.

The present invention applies in particular to a hybrid sensor whose transduction elements consist of PN junctions, generally photodiodes which transform incident photons into electrical charges.

There are mainly two types of circuits for reading the electrical charges generated in a photodiode.

Thus the FIG. 1 This illustrates a voltage follower readout circuit, where the photoelectric charge generated by the photodiode PD is stored in the junction capacitance of the photodiode and associated parasitic capacitances. The potential variation is read by a voltage follower SF in the readout circuit, whose output constitutes the pixel signal SP. This structure, known as an SFP (for "Source Follower Pixel"), has the advantage of being very simple and compact. One drawback is that the photodiode's bias can vary during exposure, causing non-linearity.

There FIG. 1 A more advanced solution called CTIA (for "Capacitive Trans-Impedance Amplifier") is proposed, which proves to be much better in terms of bias stability. One electrode of the photodiode PD, here its anode, is connected to the input of an operational amplifier AMP with a capacitor C between its other input and its output. This offers several advantages: first, the bias is much more stable thanks to the high gain of the operational amplifier; second, because the photoelectric charge generated by the photodiode is transferred to the capacitor C, the sensitivity can be adjusted by changing the amplifier's gain.

Figures 1A and 1B show that a direct electrical connection between the photodiode PD and the readout circuit is required. This interconnection is achieved through a hybrid configuration with a direct electrical contact point at each pixel of the sensor.

Figures 2A to 2D show various known technical solutions for such direct hybridization, the common principle being that for each pixel, an electrode is formed on the photodiode substrate and on that of the readout circuit. The two substrates are then placed face to face, aligned, and mechanically and electrically connected.

There FIG. 2 This demonstrates an approach involving the interposition of a PA polymer adhesive and pressure application. This pressure is maintained by removing the polymer adhesive before and during curing. While conceptually simple, this approach requires very high electrode height uniformity on both substrates and excellent tool parallelism. Furthermore, it presents a risk of failure if the stresses generated by adhesive removal are released when the sensor is exposed to adverse conditions such as high temperature, humidity, etc. For these reasons, this known approach is considered unreliable.

There FIG. 2 This demonstrates a metal fusion approach. Typically, indium electrodes are very popular due to their low melting point and high ductility at very low temperatures.

There FIG. 2 This demonstrates a more recent approach, known as "Copper-Pillar," which involves creating an electrode in the form of DC copper columns on each substrate, with one of the columns capped with SM solder material. The solder volume compensates for the uneven height of the copper columns.

Finally FIG. 2 This illustrates the most recent approach to molecular bonding, which is theoretically more reliable and reproducible. It is based on the fact that covalent bonds can be formed between two dielectric surfaces (typically silicon dioxide) in very close contact at a relatively moderate temperature (< 300 °C). This approach requires atomic-scale flatness of both contacting surfaces and extreme cleanliness, since the initial bonding involves Van der Waals forces. The implementation cost is therefore potentially high.

It should be recalled here that two key points for conventional hybridization are, on the one hand, the realization of a large number of direct electrical contacts, at the level of all the pixels of the sensor, and on the other hand, an alignment that is all the more demanding as the pitch of the pixels becomes smaller.

These difficulties are far from being resolved today, and manufacturing yields remain relatively low due to a significant proportion of defective products, particularly for high-resolution sensors. It is observed that the greatest challenge in hybridization lies in establishing reliable electrical contact between the photodiodes and the readout circuits, while precise alignment has become relatively easy to achieve thanks to advances in digital microscopes.

The present invention aims to propose a new hybridization principle enabling the resolution of all or part of the problems described above related to direct contact hybridization.

For this purpose, we propose a set of two electronic microcircuits assembled together by creating a set of electrical couplings between homologous connection points of the two microcircuits, a first microcircuit being a transducer microcircuit with at least one photoelectric transduction element by charge accumulation, and the second microcircuit being a reading microcircuit with at least one reading element of the electrical charges generated by a respective photoelectric transduction element, and the connection points comprising facing surface elements separated by a dielectric, the coupling between the two circuits being a capacitive coupling.

This circuit advantageously but optionally includes the following additional features, taken individually or in any technically compatible combinations:

* Each photoelectric capture element is a photodiode, one electrode of the photodiode being connected to a respective surface element located on the first microcircuit and its other electrode being connected to a reference potential.

* the surface element or each surface element located on the second microcircuit is connected to the input of a reading element comprising an infinite or quasi-infinite DC input impedance circuit associated with a reset switch.

* A circuit with infinite or near-infinite input impedance in direct current is a voltage follower circuit.

* the surface element or each surface element located on the second microcircuit is connected to the input of a readout element comprising a capacitive transimpedance circuit associated with a reset switch.

* the surface element or each surface element located on the second microcircuit comprises a first part connected to the input of the capacitive transimpedance circuit and a second part belonging to a feedback link, the first and second parts forming distinct capacitances with the corresponding surface element of the first microcircuit.

* The first microcircuit comprises an array of charge-accumulating photoelectric transduction elements, and the second microcircuit comprises a corresponding array of elements for reading the electrical charges generated by the respective photoelectric transduction elements.

* The assembly includes at least one direct electrical connection between the first and second microcircuits for control signals common to the different elements of the matrices.

* The first and second microcircuits each include at least one additional surface element, the two opposite surface elements providing additional capacitive coupling for control signals common to the different matrix elements.

* Additional surface elements surround the connection points.

* the additional capacitive coupling is capable of carrying a reset signal for the photoelectric transduction elements of the first microcircuit.

* The dielectric includes an adhesive for bonding the two microcircuits.

* The adhesive contains a permittivity-increasing additive.

* The dielectric comprises a layer of dielectric material assembled to the two microcircuits by molecular bonding.

* The photoelectric capture element(s) comprise an avalanche photodiode, the capacitive coupling ensuring isolation from the bias voltage of the photodiode(s).

Other aspects, objects, and advantages of the present invention will become more apparent upon reading the following detailed description of preferred embodiments thereof, given by way of non-limiting example and with reference to the accompanying drawings. In the drawings:

FIG. 1 And FIG. 1 These are schematic diagrams of a hybrid sensor pixel with two conventional reading technologies.

FIG. 2 , FIG. 2 , FIG. 2 And FIG. 2 schematically illustrate four known hybridization techniques between a substrate of transduction elements and a substrate of readout elements for such a hybrid sensor,

FIG. 3 is a diagram of the circuit of a pixel of a hybrid sensor according to a first embodiment of the invention,

FIG. 4 is a more detailed diagram of the circuit of the FIG. 3 ,

FIG. 5 is a timing diagram illustrating the different signals involved in the circuit of the FIG. 4 ,

FIG. 6 illustrates the hybridization of the two substrates in a hybrid sensor according to the invention,

FIG. 6 illustrates a variant of the hybridization of the FIG. 6 ,

FIG. 6 illustrates another form of realization of the hybridization of the FIG. 6 ,

FIG. 6 illustrates another form of realization of the hybridization of the FIG. 6 , And

FIG. 7 is a diagram of the circuit of a pixel of a hybrid sensor according to a second embodiment of the invention.

Detailed description of preferred embodiments

In what follows, two possible implementations of a hybrid image sensor according to the present invention will be described, one with the combination for each pixel of a photodiode and a readout circuit of the SFP type source follower, and the other with the combination for each pixel of a photodiode and a readout circuit of the CTIA type amplifier.

In both cases, the photodiode is reset at the beginning of each exposure either by a switching transistor or by a virtual ground generated for example by an operational amplifier.

The invention provides for a capacitive hybridization between the photodiode substrate and the readout circuit substrate, enabling the reading of potential variations on these photodiodes while ensuring the resetting of the photodiodes between two exposures. Thus, each photodiode is connected to its readout circuit without direct contact, but via a capacitor formed between two electrodes brought opposite each other and at a short distance during the assembly of the two substrates.

It is understood that thanks to the signal passing through a capacitor, there is no risk of the signal disappearing as is the case in a direct contact mode when a contact is defective, as mentioned in the introduction.

Indeed, a certain capacitance value will always exist between a photodiode and its readout circuit, and only variations in the capacitance value at a pixel can affect the readout performed on that pixel. Such uniformity defects, when they exist, are generally easy to correct through digital post-processing.

The capacitors are formed by flat electrodes of appropriate dimensions and shapes made opposite each other on the two substrates which will then be assembled, typically by usual techniques for making conductive contact plates.

We will first describe the implementation and operation of the invention with an SFP source follower read circuit.

Considering a pixel comprising a photodiode and an SFP readout circuit, it becomes clear that if a capacitor is inserted between the photodiode and the readout circuit, it is indeed possible to capture the potential variation generated by the accumulation of charge during exposure. However, it remains necessary to ensure the reverse polarization of the photodiode between two exposures (resetting the pixel), and solutions for achieving this will be described later.

There FIG. 3 This illustrates the diagram of a pixel with an SFP type readout circuit with the two substrates assembled by capacitive hybridization, namely an SPD photodiode substrate (also called transduction substrate) and an SCL readout circuit substrate.

One electrode of the PD photodiode, here its anode, is connected to an ET electrode designed to form, with an EL electrode on the readout substrate side, a capacitance called the CH hybridization capacitance, which is formed when the two substrates are joined together. The diode's cathode receives a reset signal allowing it to be reset between two exposures, as will be seen later.

On the readout circuit side, the corresponding electrode EL of the hybridization capacitor is connected to the input of a source follower amplifier SF. A recovery switch MRST (typically a simple transistor) can selectively connect the input of the SF amplifier to a reference voltage VREF. The amplifier output delivers the pixel signal SP.

When the MRST recovery switch is conducting, the EL electrode of the CH hybridization capacitor is connected to the VREF voltage (ground or a fixed voltage source), and the PD photodiode and the CH hybridization capacitor together form a rectifier circuit.

In this configuration, if an alternating or pulsed RAZ signal is applied to the cathode of the PD photodiode during this time, a quantity of charge accumulates in the CH hybridization capacitor due to the asymmetric conduction of the photodiode junction. This quantity of charge allows the photodiode to be reverse-biased to reset it to zero.

In more detail and with reference to Figs. 4 and 5, if a RAZ electrical pulse is applied to the photodiode during the application of the MRST transistor control pulse (see FIG. 5 ), it becomes forward polarized and reverse polarized after the removal of this impulse.

If the reset transistor MRST is blocked shortly after the suppression of the RAZ pulse on the substrate, the potential variation of the photodiode (voltage Vpd) under the effect of the incident light is then transmitted to the source follower SF (voltage Vin) through the hybridization capacitor CH, the voltage Vin gradually increasing (from its starting point VREF) as photoelectrically generated electrical charges are created in the photodiode PD.

The output of the SF follower thus yields a voltage Vs that reproduces this evolution, except that the starting point is equal to VREF plus the offset voltage Vos of the SF follower. This voltage Vs constitutes the output signal of pixel SP.

It is understood that the capacitive transmission gain depends on the ratio between the hybridization capacitance and the input capacitance of the voltage follower. If the input capacitance of the follower is low, the gain will be very close to 1, and this more than compensates for the variation in the value of the hybridization capacitance CH.

Now referring to the FIG. 6 In a first embodiment, the application of the reset pulse (RAZ) to the SPD substrate can be achieved via a direct connection pad (CD) to the SPD substrate. This direct connection can be made, for example, with a connecting wire on the light-exposed side of the SPD transduction substrate, as illustrated. Alternatively, one or more direct electrical contacts can be provided on the opposite sides of the SPD and SCL substrates. In this case, it is advantageous to provide multiple contacts to limit the risk of faulty connections.

Now referring to the FIG. 6 Alternatively, the RAZ signal can be routed via a CRAZ coupling capacitor between the substrate of the SPD photodiodes and the substrate of the SCL readout circuits. This capacitor is advantageously implemented with a large coupling area compared to the area of the CH capacitors existing at each pixel, so as to provide all pixels with a sufficient amount of charge to reliably perform the reset.

One advantage of this method of implementation of the FIG. 6 The key advantage is that no direct electrical connection between the two substrates is necessary. Assembling the hybrid sensor then becomes a simple gluing operation, ensuring a reasonably uniform thickness of dielectric (the adhesive itself) between the electrode pairs of the respective capacitors.

In addition, the absence of direct electrical contact between the photodiodes and their respective readout circuit reduces noise related to direct contacts, thus improving image quality.

We will now describe a variant implementation of the follower implementation described above, aimed at increasing the operating dynamics.

It is worth recalling that the reset pulse (RAZ) on the photodiode substrate during the activation of the recovery pulse (MRST) in the readout circuit initializes the photodiodes at the start of the exposure. However, if the light is too intense, some photodiodes may be completely discharged; in this case, the photoelectric charges of the saturated photodiodes will spill over into neighboring photodiodes, causing a blooming effect.

A known solution to this problem is to implement reverse-biased PN junctions (AB junctions, for "anti-glare") around the photodiodes. These reverse-biased junctions absorb the photoelectric charges that overflow from the saturated photodiodes to limit glare propagation. However, these AB junctions must be maintained in reverse bias, even during the operation of the reset pulse.

This solution has a drawback: it reduces the efficiency of the photoelectric conversion because these AB junctions absorb some of the photoelectric charge generated by the photons. However, one advantage of this solution is that it provides a logarithmic response after photodiode saturation. This logarithmic response gives the sensor a dynamic range that can exceed 120 dB.

We will now describe, with reference to the FIG. 7 the realization and operation of the invention with a CTIA type reading circuit.

In this implementation, the direct electrical contact between the photodiode and the CTIA amplifier is replaced by the FIG. 1 not by a single capacity but by two capacities CH1 and CH2.

The first capacitor CH1 plays the same role as the capacitor CH in Figs. 3 and 4. The second capacitor is added to ensure the continuity of the negative feedback necessary for the operation of the amplifier.

During the recovery of the amplifier (AMP) via the PRST switch, a negative reset pulse (RAZ) is applied to the photodiode substrate, as in the case of an SFP readout pixel. Again, this RAZ pulse establishes a reverse bias on the photodiodes, and then the amplifier's MRST switch is opened.

The negative feedback allows the bias voltage to remain constant on the photodiode and transfers the charge accumulated to the integration capacitor consisting of the capacitance CH2 during exposure.

Advantageously, the CH1 and CH2 capacitors consist of a single ET electrode on the photodiode substrate side, and two adjacent electrodes EL1 and EL2 located opposite the ET electrode on the readout substrate side.

The respective surface areas of the CH1 and CH2 capacitors are determined based on the required amplification level, among other factors.

We will now describe a method for bonding the substrate of the SPD photodiodes to the substrate of the SCL readout circuits, thus securing them together. This bonding can be achieved in various ways, for example using a polymer adhesive, molecular bonding, a solvent, and so on.

If a polymer adhesive is used, as schematically illustrated in Figs. 6A and 6B, additives can be incorporated into the adhesive to increase its electrical permittivity. High electrical permittivity increases the capacitance between the photodiodes and the readout circuits and therefore improves the signal transmission quality between the photodiodes and the readout circuits. For example, the findings of the article "Improved dielectric permittivity of NBCTO/epoxy composite films with low dielectric loss," Yanli Su, Chengdong Ba, Fei Wang, SN Applied Sciences (2019), Springer Nature Switzerland AG, can be applied.

In the case of bonding without the use of adhesive, and as illustrated in Figs. 6C and 6D, the electrodes on the substrate side of the photodiodes and/or on the substrate side of the readout circuits are coated with a thin layer of a material, typically an OX oxide, intended to form the dielectric of the capacitances created by the hybridization, the bonding then taking place by molecular bonding.

We will now describe a particular case of a hybrid sensor according to the invention, where the transduction photodiodes are of the avalanche effect type.

Avalanche photodiodes are very useful for increasing sensitivity and response time in optoelectronic systems such as single-photon sensors or LiDAR distance measurement sensors. It is known that operating these avalanche photodiodes requires a very high bias voltage, for example, over 100V for silicon-based avalanche photodiodes. This high bias voltage is potentially dangerous for readout circuits that can only withstand a few volts.

With a design that takes this into account, the compatibility between avalanche photodiodes and readout circuits is good. However, a problem arises when one or more photodiodes is/are defective, either during manufacturing or use. The high bias voltage can then reach the readout circuits via uncontrolled and unpredictable paths, causing the component to fail. The capacitive hybridization of the present invention, by preventing direct electrical contact between the photodiodes and their readout circuits, ensures electrical isolation that effectively protects the readout circuits from these high voltages.

Thus, the two embodiments above, with SFP and CTIA type readout circuits respectively, can operate with avalanche photodiodes without special adaptation.

Of course, the present invention is by no means limited to the embodiments described and illustrated. In particular, the capacitive hybridization of the invention can be implemented with different types of readout circuits, as well as with photoelectric transduction elements other than simple photodiodes.

Claims (15)

A set of two electronic microcircuits assembled together by creating a set of electrical couplings between homologous connection points of the two microcircuits, a first microcircuit being a transducer microcircuit with at least one photoelectric transduction element by charge accumulation, and the second microcircuit being a readout microcircuit with at least one element for reading the electrical charges generated by a respective photoelectric transduction element, and the connection points comprising facing surface elements separated by a dielectric, the coupling between the two circuits being a capacitive coupling. Assembly according to claim 1, characterized in that the or each photoelectric capture element is a photodiode, one electrode of the photodiode being connected to a respective surface element located on the first microcircuit and its other electrode being connected to a reference potential. Assembly according to claim 1 or 2, characterized in that the surface element or each surface element located on the second microcircuit is connected to the input of a reading element comprising an infinite or quasi-infinite DC input impedance circuit, associated with a reset switch. Assembly according to claim 3, characterized in that the circuit with infinite or quasi-infinite DC input impedance is a voltage follower circuit. Assembly according to claim 1 or 2, characterized in that the surface element or each surface element located on the second microcircuit is connected to the input of a reading element comprising a capacitive transimpedance circuit associated with a reset switch. Assembly according to claim 5, characterized in that the surface element or each surface element located on the second microcircuit comprises a first part connected to the input of the capacitive transimpedance circuit and a second part belonging to a feedback link, the first and second parts forming distinct capacitances with the corresponding surface element of the first microcircuit. Assembly according to any one of claims 1 to 6, characterized in that the first microcircuit comprises an array of charge-accumulating photoelectric transduction elements, and in that the second microcircuit comprises a homologous array of elements for reading the electrical charges generated by the respective photoelectric transduction elements. Assembly according to claim 7, characterized in that it comprises at least one direct electrical connection between the first and second microcircuits for control signals common to the different elements of the matrices. Assembly according to any one of claims 7 and 8, characterized in that the first and second microcircuits each comprise at least one additional surface element, the two opposite surface elements providing additional capacitive coupling for control signals common to the different matrix elements. Assembly according to claim 9, characterized in that the additional surface elements surround the connection points. Assembly according to one of claims 9 and 10, characterized in that the additional capacitive coupling is capable of carrying a reset signal for the photoelectric transduction elements of the first microcircuit. Assembly according to any one of claims 1 to 11, characterized in that the dielectric comprises an adhesive for joining the two microcircuits. Assembly according to claim 12, characterized in that the adhesive contains a permittivity increasing additive. Assembly according to any one of claims 1 to 10, characterized in that the dielectric comprises a layer of dielectric material assembled to the two microcircuits by molecular bonding. Assembly according to any one of claims 1 to 14, characterized in that the or each photoelectric capture element comprises an avalanche photodiode, the capacitive coupling ensuring isolation from the bias voltage of the or each photodiode.