DE3644882A1 - THERMAL DETECTOR MATRIX - Google Patents

Description

The invention relates to a thermal detector matrix and especially on a thermal infrared Detector matrix. In a thermal detector, the Energy of the absorbed incident radiation Temperature of each detector element. This temperature increase causes changes in the temperature-dependent property shafts of the detector, which are used to detect the incident Radiation can be monitored. In the present together hang particularly interesting thermal detectors contain pyroelectric detectors and / or dielectric Bolometer detectors in the form of a two-dimensional Field coupled to a semiconductor reading device is, therefore, a compact, designed as a solid  uncooled storage matrix for the generation of thermal Images and related use cases is created.

There is a growing interest in thermal Detectors in large two-dimensional fields with fine Division (for example for fields up to and over 100 × 100 elements for divisions with or below 100 µm) Achieve achievements that reach the basic limits of a thermal detector. This performance should, for example, with regard to the signal / noise Ratio of the detector matrix can be assessed. For To achieve this performance it is necessary to first do that signal that can be tapped from the matrix (the response sensitive maximize) by optimizing the matrix structure an appropriate selection of the detector material is made (i.e. selected a material with a high response quality is), the electronic reading circuit is optimized and all, for example, with the dielectric losses of the Noise sources associated with detector material as well as those in the noise sources contained in the electronic reading circuit be reduced to a minimum. Once this intoxication sources have been sufficiently reduced, however an approach to a basic one, through physical Limiting noise caused by the thermal conductivity of the structure and the effectiveness radiation detection and absorption in the detector elements is determined. This limit noise figure, too the thermal statistical noise limit is called is given by the following expression:

where G _{H is} the total thermal conductivity of an element in the matrix, η is the efficiency of radiation detection and radiation absorption in an element area, K is the Boltzmann constant and T is the absolute temperature.

It can also be shown that the electrical contribution to equivalent performance of the thermal statistical Rushing the expression

Y _H / η

is proportional, where Y _{H is} the thermal admittance of an element in the matrix, part of which is jwC _H , where w is the operating angular frequency and C _{H is} the heat capacity of the element. This expression states that a small thickness of the detector elements is required, so that wC _H in the operating frequency range is less than or equal to the thermal conductivity, ie the real part of Y _H. In practice, overall thermal conductivities of less than 0.2 w · au ^-2 K ^{-1 are} required with radiation detection and absorption efficiencies approximating 1 for element divisions at or below 100 µm and element thicknesses of less than 15 µm (preferably less than 8 µm), so that the limit of thermal statistical noise still gives usable performance at operating frequencies in the range of 10 to 50 Hz. For other frequencies, the preferred thickness changes inversely with frequency.

A pyroelectric detector that respects the equivalent noise performance has improved behavior, is described in GB patent application 21 00 058 (1982). According to the statements in this patent application, there is a thermal detector element adjacent to an absorber with a larger area attached and it will spring up one or more flexible films. The elec  trical connection is created with the help of conductors, which are laterally on the inner surfaces of this flexible Films. This construction can be in the longitudinal direction direction expanded in the form of a one-dimensional matrix be prevented, however, the requirement of lateral Expansion according to what has been said above generally the ver realization in a two-dimensional form with fine Division.

A two-dimensionally integrated detector matrix is in the US-PS 41 62 420 (1979). In this matrix however, the elements are not structured like a network. The detector level is by means of a series of metal bars (10 µm² Cross-section and 25 µm long) at a distance from one made of sili cium formed reading circuit, but with the there is an electrical connection. The usage metallic materials in this form together with the lack of network structure a device that is a relative low thermal insulation and thus a thermal Has crosstalk between adjacent elements.

So it was a problem so far, a big, two-dimensional sional integrated infrared detector matrix with fine To create division that is a satisfactory thermal Isolation and low crosstalk.

With the help of the invention, detector structures and production methods are to be created which are suitable for large two-dimensional thermal detector matrices in which high thermal insulation (low value for G _H ) is combined with high efficiency for radiation detection and absorption.

According to the invention, a thermal detector matrix is included individually arranged in a two-dimensional field  Thermal detector elements characterized in that each element on one side of a layer of elastic Material with low thermal conductivity is attached, which also serves to space the elements apart to keep, with an opposite side of the layer being a common one Electrode structure bears with every detector element is in contact and via electrically conductive tracks with high thermal resistance connections between the detector elements that each detector element with a Element electrode is provided with an inside directed surface of the detector element is in contact, each element electrode using a bonding member an input circuit of a semiconductor substrate is indicated is closed, so that the bond members serve the Layer with respect to the substrate, and that everyone Element is assigned a radiation absorber.

The bond links can be made of a lower material Thermal conductivity should be formed.

It is preferably provided that the bond members between the respective detector elements and the substrate are. As an alternative, it can be provided that the Bonds adjacent to the sides of each detector elements are attached and directly between the Extend substrate and the layer.

Furthermore, with the help of the invention, an associated manufacturing procedure created. Structures like those described above Structures can be made by full Layers of detector material are processed; you can however also using thin film technology be put.  

The improved structures and processes are suitable for dielectric bolometers and pyroelectric detector matrices, in those of single crystal or ceramic ferroelek trical oxide materials such as barium, strontium, titanate, Lead magnesium niobate, lead scandium tantalate, lead iron niobate, lead iron tungstate and lead zirconate / lead iron niobat use is made.

The complete physical separation of the detector elements of the thermal detector material in the matrix has a low thermal conductivity and a low one thermal crosstalk between the elements.

The carrier layer preferably consists of a polymer material. Polymer materials are for the backing layer most suitable, as well as for other purposes, which involve a mechanical connection or support and in addition the provision of a low thermal conductivity require. Polymer materials, usually insulating materials with a complex molecular structure are common are constructed so that they are amorphous and not crystalline Are solid, offer at ambient temperature lowest thermal conductivity values of all solid-state classes, like from the following table shows:

Material class Thermal conductivity range
(typical) W · M. ^-1 · K ^-1 metals10-400 semiconductors 1-200 crystalline inorganic solids 2-200 amorphous inorganic solids 0.8-2.0 organic polymers 0.1-0.4

A preferred class of polymeric materials for the backing layer is the polyimide family of high temperature polymers. Examples can be found among those materials which have so far been developed on a large scale as passivation and dielectric intermediate layers for use on integrated silicon circuits. A specific example of a polyimide material that can be successfully given here is the polyimig material from Hitachi, designated "PIQ". The use of a carrier film made of polyimide material, which is tough and flexible and has a high glass transition temperature with good adhesion to metal and oxide surfaces, is complete with the ferroelectric materials mentioned above and with the flip-chip solder bonding process, the process preferred for the present purpose compatible. Polyimide films are applied by means of spin coating and are typically cured at temperatures between 300 and 350 ° C. The layer thicknesses over the range of ∼0.2 to 0.3 µm are possible in practice, with coatings in the submicron range being preferred. The difference in the coefficient of thermal expansion of the polyimide film and the ferroelectric oxide materials that form the detector elements (∼50 ppm · K ^-1 of <10 ppm · K ^-1 ) puts the carrier film under a low positive voltage after cooling from the hardening stage. This keeps the film taut after patterning and results in a low microphone effect. Other possible polymer materials for use as a carrier film contain polyamides and polyurethane (including UV-curable types) and polymers separated from the vapor phase (for example parylene).

The use of a completely single item structure built on a polymer material Carrier layer that is under a low tension, ensures a very low microphone effect. A  certain piezoelectric activity is both in the pyroelectric as well as in the dielectric bolometer materials under the influence of the applied preload field exists, and the detectors must be so designed be that this undesirable response to a Minimum is reduced.

The radiation absorber structure mentioned above can be made from a thin resistive metal film with one attached to it Free space impedance matched impedance (377Ω10) be formed, which also serves as the common electrode. Such a film absorbs 50% of the incoming broadband radiation. To increase the absorption efficiency over a limited waveband can prevent reflection changing layers are added. Also an active one Reflective metal electrode layer on top of the other Side of the detector element can be used together with the Knowledge of the optical infrared properties of the materials in the structure and control of the thickness of the Detector element can be applied to the absorption to further increase efficiency in the band of interest. Efficiencies from 0.6 to 0.8 in the 8-14 µm infrared band are achievable with such multilayer structures. One before the absorber layer structure comprises a metal black like platinum black that is close to one Absorption coefficient in a structure with very low thermal mass offers.

The absorbers can have an area that the corresponds to the respective detector elements, but the Absorber and detector surfaces in a networked detector matrix structures can be changed independently. The independent nature of these two structures can do this be used, a high radiation detection and absorption efficiency combined with a high degree  thermal insulation between the detector elements to achieve.

Preferred electrical materials especially for Verbin tracks between the elements and to the contact surfaces of the elements contain metals, metal alloys, Cermete and amorphous metal alloys with high heat resistance; they have good liability to others Materials in the structure and can be used for formation narrow paths are structured. Examples included anodized chrome, metallic nickel-chrome alloy films, chromium-silica-cermet films and amorphous Nickel-indium metal films. Such films with less Thickness (e.g. 10 to 30 µm) and narrow width (for example 2 to 10 µm) can be required using a variety of thermal applications procedures, including the Filament and electron beam evaporation or by Atomization. Sputtering processes and especially magnetron Atomization processes are preferred. To apply An electrode is made from metal black like platinum black one of the above materials with a coating made of an inert metal with high electrical conductivity required like gold. The inert metal with high conductivity ability can depend on the connecting tracks of the element the application of the metal black are removed so the associated thermal tributary is eliminated. Through holes in the polymer carrier film over the Elements are desirable so that good thermal contact guaranteed between the absorber and the detector element becomes. If a thin matched in terms of impedance Film-absorber structure is used, the polymer carrier film also as a reflection preventing layer for the Serve absorber.  

The invention will now be described by way of example with reference to the drawing explained. It shows

Fig. 1 shows a section of an embodiment of an integrated thermal detector array according to the invention,

FIG. 2 shows a detailed section of an image element of the matrix from FIG. 1, the structure of the bonding element being shown exactly,

Fig. 3 is a sectional view of an integrated thermal detector array in accordance with another embodiment of the invention,

FIGS. 4 and 5 is a section and a plan view of a preferred embodiment of the matrix shown in the previous figure,

FIGS. 6 and 7 are schematic sections of the matrices according to the preceding figures, in which individual units of the construction and assembly are shown in a production of a material body,

Fig. 8 and 9 are schematic steps of a pre-formed part and the matrix in the manufacture of using the thin film technology.

For a better understanding of the invention, Ausfüh Example with reference to the drawings explained.  

An integrated thermal detector matrix is shown in FIG . It consists of a large number of thermal detector elements 1 , each of which is held individually at a distance from one another in a matrix position by means of an elastic carrier layer 3 made of a material with low thermal conductivity. The carrier layer provides sufficient hold and mechanical strength for the entire arrangement. On an inwardly facing surface of the layer 3 extends a common electrode structure 5 , which produces a connection between the elements with the aid of electrically conductive connecting members 7 with high thermal resistance. The other surface of each element 1 carries a single element electrode 9 . The held matrix structure is arranged over a semiconductor substrate 11 , and bonding elements 13 , which also have a supporting function and make electrical connections, are inserted between each element 1 and the substrate 11 . These bonding elements produce the electrical connection between the element electrodes 9 and corresponding input contact areas 15 of interface reading circuits which are integrated in the semiconductor substrate 11 . According to each detector element 1 , a block 17 of heat-absorbing material is placed on the other surface of the layer 3 . There is thermal contact between each element 1 and the absorber block 17 , and effective heat conduction is created through the carrier (layer) or through metal-filled openings in the carrier (web).

As shown in Fig. 2, each bond member 13 is a composite structure which is placed below each element 1 . In this way, the conflicting requirements of the electrical connection and the thermal insulation are combined. This composite structure can be realized, for example, that immediately below the detector element 1, an elevation 19 of a material with low thermal conductivity, for example a polymer, is formed and on the elevation 19 a thin metal sheet 21 is attached with high thermal resistance. The metallized polymer elevation 19 is then electrically connected by bonding to the input contact surface 15 of the substrate 11, for example by providing a solder bond connection 23 and a surface metallization 25, 27 . This structure can also be used in inverted form, ie the solder bond connection 23 is attached adjacent to the detector element 1 and the polymer bump 19 adjacent to the substrate 11 . However, this structure is less preferred because the solder bond 23 then acts as a thermal load on the detector element 1 and accordingly reduces the response behavior.

A number of polymer materials can be selected to form polymer bump 19 . Polyimide materials are particularly suitable for the inverted shape of the bump structure, in which the bump 19 is arranged on the substrate 11 . Polyimide materials are less suitable for the original bump structure because the curing cycle involved can pose some difficulties in the manufacturing process. For the formation of the structure of FIG. 2, a Novolac photo resist is preferably used, which can be structured and then flow dried to form an elevation 19 with a rounded edge profile to facilitate the subsequent metallization. To structure the Novolac resin and to stabilize the bump dimensions from the metallization and the final solder bonding operation, it is desirable to carry out a radiation stabilization treatment (UV, X-ray or electron beam) or a chemical stabilization treatment (for example immersion in an acidic formaldehyde solution).

The assumption of a point-bonded structure in which the bond members 13 lie below each element 1 means that the element itself can fill up a large part of the available pixel area. It can be shown that with an optimized compromise between the pixel fill factor (ie the radiation detection efficiency) and the thermal crosstalk between neighboring elements 1 for equal areas of the absorber blocks 17 and the elements 1 with a ratio of the element width to the gap width of 4: 1 corresponding to a radiation detection efficiency of 0.64, the use of an expanded platinum black collector area (see FIG. 4) and / or associated collector structures (XAC) with an enlarged area, a further considerable gain in the detection efficiency can be achieved without impairing the thermal insulation. With the aid of such means, a detection and absorption efficiency approximated to 0.9 can be achieved, in particular in structures with or with a pitch of less than 100 μm.

In Fig. 3, an alternative embodiment of a matrix structure is illustrated. In this example, the thermal insulation of the element 1 from the reading circuit substrate (the main heat sink in each structure) and between neighboring elements is achieved simultaneously with the aid of a polymer carrier tape 3 around each element 1 . The individual elements 1 of this structure are smaller than in the case of the point-bonded structure ( FIG. 1), and the polymer carrier web 3 is correspondingly longer. The bond members 13 , which produce the electrical connections from the element electrodes 9 to the read circuit 15 , are introduced on the web 3 , and it is no longer essential to attach additional heat barrier structures (bumps) in the bond members 13 , as used in the point bond structure has been. With this structure, the overall thermal conductivity is a compromise in terms of radiation detection efficiency (in the event that element 1 and absorber block 17 have the same area), and an optimal element diameter is 0.6 of the pixel width, which is a pixel fill factor ( or radiation detection efficiency) of about 0.35. The total and lateral thermal conductivity values for this polymer membrane structure are much lower for a given pitch and for practical structure dimensions than for the dot bond structure of Fig. 1. This can finally result in better matrix performance, provided that the radiation detection efficiency increases can be.

A preferred embodiment of the two-dimensional thermal detector matrix attached to a membrane contains a collector 17 ' made of platinum black with an enlarged area (PB-XAC), as shown in FIG. 4. It should be particularly noted that this preferred structure includes trimming the corners of the detector elements, resulting in octagonal and non-square elements. This leads to an extension of the thin-film metallization path 29 , which connects each element electrode 9 with each bonding element to the reading circuit 19 , which results in a further increase in thermal insulation with a minimal loss of absorber area. This thin-film web 29 runs from the element electrode 9 via the polymer insulation 31 to the point where the bond members 13 are arranged. The use of four common electrode connections 7 should also be mentioned. Although two connections per element plus additional connections at the ends of each element row are sufficient for the complete joint connection, a certain redundancy is useful in the fine path geometries that are desired in these fine pitch structures. The use of three or four common connections 7 is therefore desirable. As a preferred method for producing the bond members 13 between the matrix and the electronic reading circuit 15 , the flip-chip solder bonding method has been used again. This method is particularly suitable for the membrane structure, since the solder bonding method does not involve the application of mechanical pressure, as is required for all solid phase bonding methods, for example the indium bonding method, which would be incompatible with the attachment of the bonding elements on a flexible polymer film . Rather, the solder bonding process is based on the natural wetting effect of molten solder on the wettable metallization 25, 27 , which is provided on the two components used to form the bond members 13 . The aspect ratio of the solder bond member allows a spacing of each thermal detector element 1 from the substrate 11 of the reading circuit to be maintained at a bond element height compatible with the diameter of the block that can be wetted by the solder, provided that the element thickness is less than the height of the solder bond element. This requirement for the geometry of the bond elements is compatible with the requirement for very small element thicknesses, as has already been mentioned above.

Matrix production

There will now be examples of manufacturing processes thermal detector matrices with the structures described above considered. These methods do contain certain novel ones Characteristics, but they are not necessarily the only ones Procedures or consequences of procedural steps with which such matrices can be made. It will Described methods with the help of detector matrices  massive bodies made of ferroelectric material or applied thin ferroelectric films will. Special features include procedures for selek tive thinning and to produce the network structure of the Individual elements, new structures to create the mechanical mounting of the matrices and methods for Alignment of the hybrid devices.

Matrices made of diluted solid material bodies

The structures of two-dimensional thermal detector matrices, in which the point bonding and the attachment of the elements to a polymer membrane are used and the production is made from dilute solid material bodies, are shown in FIGS . 6 and 7. These figures show the edges of the matrix structure as well as parts of the actual matrix area. The edges of the structure consist of relatively thick "picture frames" 33 without a network structure, which extend around the entire detector matrix. This edge forms a sufficiently rigid mechanical holding frame for the polymer membrane 3 . It also forms an area of the device that can be contacted during assembly, and also provides a location for the attachment of one or more rows of solder bond sites 35 which are substantially larger in diameter than the solder bond sites 23 of the actual matrix. These solder bond sites 35 with a larger diameter are provided so that the hybrid structure is automatically aligned during bonding, in particular so that the matrix of the bonding connections 23 of the individual matrix elements leading to the substrate 11 and having a small diameter is aligned.

A possible sequence of production steps for the matrix devices of FIGS. 6 and 7 comprising aligning, cutting, lapping and polishing of the solid body of ferroelectric material, either in single crystal or polycrystalline ceramic mold for producing a plate with parallel sides and with the for the thickness 33 of the device which does not have a network structure. The area from which the matrix having the network structure is to be produced is then selectively thinned to the final thickness of the device. Suitable selective dilution methods can include various wet and dry etching processes, for example optically enhanced chemical etching or ion beam etching, which depends on the material and its crystal shape. The ion beam etching has the advantage that it is a relatively universally applicable etching process, so that it is preferred in many cases. The areas that should not be thinned are appropriately masked against the etching used, corresponding to the respective etching process, for example by means of a metal foil mask or a photoresist layer.

The various layers required for the surface 37 of the device exposed to the incident radiation are then applied in the selectively thinned depressions. They usually contain an etching stop layer for the subsequent process for forming the network structure, the polymer membrane layer 3 , the layers 5 forming the common electrode and the infrared absorber structure 17 . The plate is then turned over and the element electrodes 9 and the masking layers required to form the network structure are added. The network structure is preferably formed by means of an anisotropic etching process, for example by means of ion beam etching or etching with reactive ions. Masking materials and layers of etchant can be appropriately selected in accordance with the materials to be etched, the etching process and the present geometry. Below the support edge 33 of the device, a group of recesses 39 can be etched simultaneously with the formation of the network structure of the elements of the matrix, these recesses later receiving the alignment solder bonds 35 with a larger diameter. Etched depressions 39 are particularly suitable for the structure attached to the membrane, as can be seen from FIG. 7. If the solder bond connections 23, 35 are placed on the reading half of the fixed hybrid device which is formed by the silicon chip, then these depressions 39 can also contribute to mechanical alignment of the components before the solder is melted, in order to complete the bonding of the hybrid device. As soon as the formation of the network structure has ended, the etch stop layer serves to protect the underlying polymer membrane 3 . After it has served its purpose, the etch stop layer can be removed if desired.

After the network structure has been formed, the various layers that are required on the element connection side of the structure are added. These layers contain the polymer insulating layer 31 , the element electrode 9 , the polymer elevation layers 19 , the connecting tracks 21 from the elements for the reading circuit and the moistenable metallization contact surfaces 25, 27 for receiving the solder bond connections 23 . A corresponding group of humidified metallization contact surfaces 27 are provided on the input contact surfaces 15 of the reading circuit. The solder bond connections 23 are formed either on one or on both components of the hybrid device.

Thin film devices

The ferroelectric materials which form the sensitive elements of the thermal detector matrix can expediently also be applied directly in the form of a thin layer with the desired thickness of the active elements. This may avoid expensive crystal growth, ceramic fabrication, alignment, cutting, lapping, and polishing procedures that must be used when using solid material bodies for these devices. Methods of applying the thin layer of such ferromagnetic materials include reactive sputtering of composite targets, chemical application of vapor-phase metal-organic compositions (MOCVD), and on-site decomposition of metal-organic layers applied as precursors (e.g. layers produced by means of the sol-gel process or in-situ hydrolysis of spin-on metal-organic complex solutions). Such layers can be polycrystalline and have various preferred crystallographic orientations; but they can also be monocrystalline depending on the conditions of the application and / or the decomposition. Particularly important parameters in this context are the application or decomposition temperature and the type of substrate on which the layers are formed. These processes for such relatively complex materials as the above-mentioned ferroelectric oxides usually involve deposition or decomposition temperatures above 400 ° C, especially when highly crystalline layers with preferred orientation or monocrystalline layers are desired. For this reason, direct application to a suitable thermal insulation structure that is already in place on the reader is usually difficult. The application of the layer is therefore preferably carried out on a suitable intermediate substrate. A substrate 41 with an applied ferroelectric layer 43 and a separating layer 45 is shown in FIG. 8. The selection of a suitable heat-resistant substrate material can remove the above-mentioned limitation regarding the temperature of the application process and at the same time open up the prospect of controlling the layer structure. Suitable substrate materials contain amorphous materials (e.g. molten quartz) and crystalline oxides as well as other inorganic composite materials. Monocrystalline materials (for example spinel, sapphire, magnesium oxide, cubic zirconium dioxide, sodium chloride and lithium fluoride) are particularly interesting because they enable the epitaxial growth of monocrystalline ferroelectric layers if the lattice parameters or the lattice spacings are suitably adapted. Correctly aligned monocrystalline ferroelectric layers can result in higher quality figures of the material (e.g. due to a lower dielectric loss than the equivalent polycrystalline layer).

The separating layer 45 shown in FIG. 8 may be necessary in order to prevent a chemical interaction of the grown layer with the substrate and / or to contribute to a core formation of the layer. The separating layer 45 can in turn be applied to the substrate 41 in such a way that a preferred orientation or an epitaxial monocrystalline layer results. The preferred orientation or the monocrystalline nature of this separating layer can continue in the ferroelectric layer grown thereon. The separation layer 45 can also be used to achieve a gradual transition between the lattice parameter of the substrate and that of the ferroelectric layer 43 , where epitaxial growth is carried out. An example of a separating layer 45 is metallic platinum, which can be applied by sputtering with a preferred orientation on molten quartz or as an epitaxial layer on monocrystalline magnesium oxide.

After a ferroelectric thin layer 43 having the desired thickness has been formed on the substrate 41 , the structure will then be processed in a very similar manner to the structure formed from a solid material body, as shown in FIG. 9. The substrate layer 41 is etched off in the active region of the matrix in a manner suitable for the selected substrate material, but is retained on the circumference of the structure, so that the thicker, non-network-holding holding frame 33 is created for the polymer membrane layer 3 , which carries the detector matrix . The separation layer 45 can be contained in the matrix structure or removed, depending on what is required. Otherwise, the processing takes place in the same way as described above, and the result is a spatially similar structure.

Further structural variants which can be derived from the structures described above are within the scope of the invention. Particular examples include the use of alternative arrangements for concentrating and / or collecting the incident infrared radiation on the detector elements 1 . Refractive or reflective optical structures including small optical immersion lenses for the individual elements as well as flat faceted or curved reflective funnel structures are included. Such structures can be produced, for example, by photochemical or anisotropic chemical etching of monocrystalline silicon dioxide with the subsequent application of suitable reflection-preventing or metallizing coatings. Such concentration structures can be assembled with the hydride device and precisely aligned to them by using the flip-chip solder bonding process.

Claims (6)

1. Thermal detector matrix with individual thermal detector elements arranged in a two-dimensional field, characterized in that each element ( 1 ) is attached to one side of a layer ( 3 ) of elastic material with low thermal conductivity, which also serves to separate the elements to keep each other, with an opposite side of the layer carries a common electrode structure ( 5 ) which is in contact with each detector element and via electrically conductive tracks ( 7 ) with high thermal resistance connections between the detector elements that each detector element ( 1 ) with Element electrode ( 9 ) is provided which is in contact with an inwardly facing surface of the detector element, each element electrode ( 9 ) being connected to an input circuit ( 15 ) of a semiconductor substrate ( 11 ) with the aid of a bonding element ( 13 ), so that the Bonding members ( 13 ) serve the layer ( 3 ) with respect to the substrate ( 11 ) and that each element ( 1 ) is assigned a radiation absorber ( 17 ). 2. Detector matrix according to claim 1, characterized in that the bonding members ( 13 ) are formed from a material with low thermal conductivity. 3. Detector matrix according to claim 1 or 2, characterized in that the bonding elements ( 13 ) are arranged between the respective detector elements ( 1 ) and the substrate ( 11 ). 4. Detector matrix according to claim 1 or 2, characterized in that the bonding members ( 13 ) are attached adjacent to the sides of the respective detector elements ( 1 ) and extend directly between the substrate ( 11 ) and the layer ( 3 ). 5. Thermal detector essentially according to the previous one lying description with reference to each of the accompanying drawings. 6. Method of making a thermal detector matrix essentially according to the present description.