Thin film transistors

[0002] This application is a continuation-in-part of application Ser. No. 09/137,920, filed Aug. 20, 1998.

[0003] The invention relates to thin film transistors (TFTs), and more particularly to TFT integrated circuits (ICs) with new designs to facilitate interconnection.

[0004] Semiconductor integrated circuits are ubiquitous in commercial electronics. ICs for electronic logic and memory devices are nearly always silicon based devices and ICs for photonic applications are typically III-V and II-VI based devices. Transistors, both metal-oxide-semiconductor (MOS) and bipolar, are made by creating impurity regions in the semiconductor substrate, and forming polysilicon or metal electrodes on the surface of the semiconductor to contact or interact with the underlying semiconductor regions. Both planar and mesa substrate configurations are used. In the manufacture of these devices, the critical nature of the semiconductor substrate, and the formation of substrate layers and electrical contacts to the substrate, dictate that the fabrication sequence begins with processing the semiconductor substrate. Subsequently in the process, after the arrays of sources, drains and gates, or emitters, bases and collectors, are formed, contacts are made to these elements to form arrays of transistors. The final phase of the fabrication is to interconnect the arrays of transistors. Thus the typical semiconductor IC is built using many layers, beginning with an active semiconductor substrate. For example, in the manufacture of a typical silicon MOS transistor, the layers may comprise a field oxide, a gate dielectric, a gate metal layer, a first interlevel dielectric layer, a first level metal interconnection layer, a second interlevel dielectric layer, a second level metal layer, and an insulating capping layer. In the essential sequence of semiconductor IC manufacture to date, the metal interconnection layers are formed last. This sequence is advantageous as long as the critical fabrication steps are performed early in the process. This factor reduces device cost by having the highest incidence of processing failures occur early in the manufacturing sequence. However, a significant drawback to this sequence is that the elements formed by these early critical steps are vulnerable to processing conditions used in later steps to complete the device. This is especially true for the semiconductor substrate, the impurity regions in the substrate, and the interfaces between the semiconductor device elements and insulating and contact layers. These portions of the device are particularly sensitive to heat. Thus e.g. the techniques used to deposit metal interconnect layers and interlevel dielectrics have been developed with careful restrictions on thermal processing. Accordingly, the typical IC manufacturing process that has evolved over time is seriously constrained due to this well known thermal susceptibility.

[0005] As IC device density increases, interconnection issues become more prominent. While individual transistor design has dominated IC technology to date, device density is becoming so large that critical steps in future IC manufacturing are likely to involve metallization and interconnect operations. In that event the least costly processes will be those that can form the interconnections first. Moreover, even in relatively low cost IC technologies, where high transistor performance is not required, interconnection technology and robust interconnection strategies may become dominant design issues.

[0006] In recent years, IC technologies have been proposed that use organic semiconductor transistors. The chief attractions of such circuits stem from the anticipated ease of processing and compatibility with flexible substrates. These advantages are expected to translate into a low-cost IC technology suitable for applications such as smart cards, electronic tags, and displays.

[0007] TFT devices are described in F. Garnier et al., Science, Vol. 265, pp. 1684-1686; H. Koezuka et al., Applied Physics Letters, Vol. 62 (15), pp. 1794-1796; H. Fuchigami et al., Applied Physics Letters, Vol. 63 (10), pp. 1372-1374; G. Horowitz et al., J. Applied Physics, Vol. 70(1), pp. 469-475; and G. Horowitz et al., Synthetic Metals, Vol. 42-43, pp. 1127-1130. The devices described in these references are based on polymers or oligomers as the active materials, in contrast with the amorphous silicon TFT structures that were developed earlier. The devices are typically field effect transistors (FETs). Polymer active devices have significant advantages over inorganic semiconductor TFTs in terms of simplicity of processing and resultant low cost. They are also compatible with polymer substrates used widely for interconnect substrates. Polymer TFTs are potentially flexible, and polymer TFT ICs can be formed directly on flexible printed circuit boards. They also have compatible coefficients of thermal expansion. While metal-insulator-semiconductor (MIS) FET devices are most likely to find widespread commercial applications, TFT devices that utilize both p-type and n-type organic active materials are also known. See e.g., U.S. Pat. No. 5,315,129. S. Miyauchi et al., Synthetic Metals, 41-43 (1991), pp. 1155-1158, disclose a junction FET that comprises a layer of p-type polythiophene on n-type silicon.

[0008] Recent advances in polymer based TFT devices are described in U.S. Pat. No. 5,596,208, issued May 10, 1996, U.S. Pat. No. 5,625,199, issued Apr. 29, 1997, and U.S. Pat. No. 5,574,291, issued Nov. 12, 1996, all of which are incorporated herein by reference, especially for descriptions of useful materials. With the development of both n-type and p-type active polymer materials, as described in these patents, complementary ICs can be readily implemented, as detailed particularly in U.S. Pat. No. 5,625,199.

[0009] For the purpose of definition the term organic semiconductor is intended to define that category of materials which contain a substantial amount of carbon in combination with other elements, or that comprises an allotrope of elemental carbon, and exhibits charge carrier mobility of at least 10⁻³cm²/V.s at room temperature (20° C.). Organic semiconductors of interest for TFTs typically have conductivity less than about 1 S/cm at 20° C.

[0010] Although the new TFT devices represent a significant departure from the semiconductor IC technology of the past, the configurations of these devices, and the overall fabrication approach, follow closely those used for silicon IC fabrication. For example, even though the substrate in these TFT devices is as critical an element as in traditional IC structures, either from the standpoint of performance or manufacturing yield, these IC devices are consistently produced by forming the active devices in a transistor array, and then forming the interconnections.

[0011] We have developed a radically new approach to integrated circuit fabrication in which the interconnections are formed prior to forming the active transistors. The IC devices resulting from this sequence have inverted structures, with the interconnections buried next to the substrate and the active elements on top. This approach follows an analysis of transistor fabrication sequence, and the realization that the use of organic semiconductor active materials removes critical constraints present in traditional semiconductor IC manufacturing. In particular, the thermal constraints mentioned earlier are removed, allowing a wide choice of processing conditions for fabricating the interconnection levels. Moreover, if interconnection strategies, performance and yield become dominant with new low cost TFT technologies, this invention allows the interconnection phase of TFT IC manufacture to occur early in the sequence, thus affording the potential for higher yields and lower cost.

[0012] FIGS. 1-17 are schematic representations of process steps useful for producing one embodiment of an inverted IC according to the invention;

[0013] FIG. 18 is a schematic view of a second embodiment of an inverted IC; and

[0014] FIGS. 19 and 20 are schematic views of an inverted TFT of the invention built on a flexible substrate.

[0015] With reference to FIG. 1, a portion of an IC substrate is shown at 11. A single inverted TFT will be illustrated for simplicity, but it will be understood that the single device is representative of a large integrated array of devices. Also, the features shown in the figures herein are not to scale.

[0016] The substrate is preferably an insulating material such as glass or a polymer. It may be rigid or flexible, and it may comprise a standard printed circuit substrate of expoxy or polyimide. Alternatively it may be silicon on which an insulating layer of SiO₂is grown or deposited. The first level metal is shown at 12. In this inverted structure this level is referred to as the first level because it is formed first but, as will be appreciated by those skilled in the art, it corresponds to the second level metallization in traditional structures. The metal may be any of a variety of conductive materials. The common choice in standard IC technology is alumninum. However, due to the nature of the structures described here the choice of conductive material can be made from a larger universe than is usually considered, including the standard materials, i.e. aluminum, TiPdAu, TiPtAu, TaN_x, TiN_x, etc., as well as non-traditional choices most notably, copper and conductive polymers such as polyaniline and metal containing polymer inks. The use of polymer conductors may be favored in applications where a degree of flexibility is desired. The choice of deposition techniques is also wider since the structure at this stage in the processing, as contrasted with traditional IC processing at this stage, have no thermally sensitive components. Thus this deposition step, as well as subsequent deposition and etching steps used for forming the two level or multi-level metallization interconnections, may involve significant substrate heating if that is otherwise convenient and cost effective. Thus the metal layer can be evaporated, or sputtered. The thickness of the metal layer can vary widely, but will typically be in the range 0.05 to 2 μm.

[0017] The next step, represented by FIG. 2, is to pattern the first level metallization using a lithographic mask 13, which is typically a photolithographic mask, but may be formed using other forms of lithography. Other masking steps, to be described below, may also utilize these alternative lithography technologies. The first metal layer is then patterned by standard etching, e.g. plasma or RIE etching, to produce the pattern of metal runners 14 as shown in FIG. 3.

[0018] With a wide choice of conductive materials available, it may be useful, in applications where the interconnect density is not large, to print the circuit directly, using screen printing, stenciling, ink jet printing or a similar technique.

[0019] With reference to FIG. 4, the first interlevel dielectric 15 is formed over the first level metal pattern as shown. The interlevel dielectrics in the structures according to the invention may be of a variety of insulating materials such as spin on glass (SOG), or Si₃N₄or SiO₂deposited by CVD for example. In the TFT structures described here, it is expected that the use of polymer materials wherever they can be effective will be desirable, both from the standpoint of processing simplicity and cost, and also to produce IC structures that tolerate stram, i.e. are somewhat flexible. Accordingly, for such applications the use of polyimide or similar organic polymer insulators is recommended. A suitable material is a polyimide supplied by Nissan Chemical Company under the designation RN-812. This material can easily be produced in layers with 0.1-1 μm thickness, which have desirable insulating properties. The application technique for organic insulators is typically spin coating or solution casting. Some inorganic insulators, notably spin-on-glass, also share the property of convenient application. In some applications, where fine pattern dimensions are not required, the dielectric layer may be applied as a patterned layer, already containing the interlevel windows.

[0020] The interlevel dielectric is then masked if required with patterned mask 16 as shown in FIG. 5, and the portion of dielectric layer 15 exposed by the opening 17 in the resist is etched to form a window to interconnect the first and second levels. The mask opening is aligned to metal runner 14 in the first level interconnection pattern. A single interlevel interconnection is shown for simplicity, but a typical IC will have many such interlevel interconnections. These interlevel interconnections are standard, and techniques for forming the interlevel windows are well known. For example, if the dielectric layer is SiO₂the windows may be formed by plasma etching or RIE. The resulting structure is shown in FIG. 6, with interlevel window 18 formed in the dielectric layer 15. Alternatively, interlevel windows or vias can be made directly using a photodefinable polymer dielectric such as polyimide, or, if polymer material is used for the interlevel dielectric the vias can be made using laser processing.

[0021] The second level metal, 19, is deposited over the first interlevel dielectric 15 as shown in FIG. 7. The second level metal may be the same as, or may be different from, the first level metal. The second level metal is patterned in a manner similar to the first level using mask 21 as shown in FIGS. 8 and 9. One of the runners 22 in the second level metallization interconnects as shown with the first level metallization 14 at interlevel window 18.

[0022] The next step is to form the second interlevel dielectric 23 as shown in FIG. 10. This layer may be formed in a manner similar to layer 15, and interlevel dielectric 23 is also provided with through holes or windows (not shown) for interlevel interconnections between the second level and the gate level to be formed next.

[0023] The gate level metal, usually the first level metal in a traditional structure, and usually of polysilicon, is formed late in the sequence of the invention, and may comprise a wide variety of metals. The usual requirement that the gate level metal be relatively refractory to withstand the conventional implantation drive steps is eliminated in the process of the invention, so the gate material can be selected from many materials, even aluminum or copper. However, the art has extensive experience with silicon gates insulated with grown SiO₂. Tantalum gates insulated with TaN or TiN may also be convenient. The gate metal layer 24 is shown in FIG. 11 deposited over the second interlevel dielectric layer 23 and into the windows (not shown)that interconnect gates to the first level metal. The gate metal layer is then patterned (FIG. 12) by conventional lithography to form gate structures 25. Conducting polymers are also suitable for the gate metal and are especially compatible with other elements in the structures described here. Gold and indium tin oxide (ITO) are also useful gate electrode materials.

[0024] The gate dielectric 26 is then formed over the structure as shown in FIG. 13. The gate dielectric may be of a conventional oxide or nitride as indicated above, or may be SOG or an organic insulator such as polyimide that can be formed conveniently by spin-on techniques. An example of such a material that has been used successfully in this application is pre-imidized polyimide, supplied by Nissan Chemical Company under the designation SE-1180. This material can be spun on at 4000 RPM and cured at 120° C. for 2 hours to produce a coating with a thickness of 70 nm. If desired, the gate material may be polysilicon, and the gate dielectric grown as a surface layer over the polysilicon in which case the gate dielectric layer 26 would not cover the entire second interlevel dielectric as it appears in FIG. 11.

[0025] The source/drain contact layer 27 is then deposited over the structure as shown in FIG. 14, and is then patterned using conventional lithographic mask 28, shown in FIG. 15, to define the source electrode 29 and drain electrode 30 as shown in FIG. 16. Alternatively, the source and drain can be formed using known additive technology, e.g. by printing using conductive polymer inks. The source and drain electrode materials may be polysilicon or any of a number of metal conductors, or may be organic conductors such as polyaniline. For display applications the electrodes may be indium tin oxide. In the work that resulted in this invention the source and drain electrodes were gold.

[0026] The final essential step in the process of the invention, which is the first step in the traditional FET process, is illustrated in FIG. 17 and is the formation of the active semiconductor body 31 in which the field effect is realized, and in which the FET channel extends between source 29 and drain 30. In this invention, the active material is preferably an organic semiconductor, but may also be an inorganic TFT material such as amorphous silicon, polysilicon, CdSe, CdS, ZnS, TiO₂, ZnO, Cu₂S. As an example of the use of an inorganic material, undoped α-Si can be deposited by plasma-enhanced chemical vapor deposition (PE-CVD) or RF sputtering.

[0027] A wide variety of organic semiconductors have now been developed for TFT devices. Among these are:

[0028] i. perylenetetracarboxylic dianhydride (PTCDA), the imide derivative of PTCDA;

[0029] ii. napthalenetetracarboxylic dianhydride (NTCDA);

[0030] iii. fluorinated copper pthalocyanine;

[0031] iv. α-sexithiophene;

[0032] v. p,p′-diaminobisphenyls in polymer matrices;

[0033] vi. tetracene or pentacene, or end substituted derivatives thereof;

[0034] vii. oligomers of thiophene with the degree of oligomerization ≧4 and ≦8, linked via the 2- and 5-carbons;

[0035] viii. alternating co-oligomers of thienylene and vinylene, with thiophenes as terminal groups and 3-6 thiophene rings, linked via their 2- and 5-carbons;

[0036] ix. linear dimers and trimers of benzo[1, 2-b: 4, 5-b′] dithiophene;

[0037] x. oligomers of v. and vi. with substituents (e.g., alkyl substituents with 1-20 carbons) on the 4- or 5-carbon of the end thiophenes;

[0038] xi. regioregular poly(thiophene)s.

[0039] Both p- and n-type materials are contained in this list and can be combined as needed for complementary ICs. For complementary MOS (CMOS) devices, combinations of organic and inorganic semiconductors can be used as described in my U.S. Pat. No. 5,625,199 issued Apr. 29, 1997. In some applications the higher mobility obtained in many inorganic semiconductors, such as CdSe, zinc sulfide, polysilicon, and others may be advantageous and used alone or in combinations. An example of such an application is a so-called RF (radio frequency) identification tag.

[0040] The TFT structure described and produced by the foregoing sequence of steps is but one form of TFT to which the invention can be applied. An alternative is shown in FIG. 18. This device is a modified form of J-FET with n-type (or p-type) layer 41 and n-type (or p-type) layer 42 together forming a p-n junction. The gate 43 controls the pinch-off of the channel between source 45 and drain 46. The first level metal is shown at 47 and the second level metal 48 are essentially the same as described in connection with the FET of FIG. 17.

[0041] As mentioned above, for simplicity the interconnections between the gate level, or the source and drain electrode level, and the interconnect levels are not shown but are standard. For example, an interconnection between gate 43 and first level metal runner 47 of FIG. 18 would be made via a window through layer 42, and through a capping insulating layer, if present. Contact with the gate 43 may be made directly, via an extension of layer 43 (in the dimension normal to the plane of the figure) or through an opening in a capping insulating layer. In the case of the device of FIG. 17, interlevel interconnections between the gates and the second level metal runners (or the first level runner in a single metallization level IC) would be made via windows in the second interlevel dielectric.

[0042] The devices shown in FIGS. 17 and 18 are representative of the generic category of field effect transistors, and demonstrate the principle of the invention that can be applied to any form of FET device, namely that the interconnect levels are formed first and the transistor last.

[0043] As indicated earlier, the features in the figures are not necessarily to scale. The dimensions of the active devices, i.e. the TFTs, can be made very small using fine line techniques. In particular, the source-to-drain spacing can be 5 nm or less. At these small dimensions a single polymer chain, or a few organic molecules span the source-to-drain distance. With such an IC technology, it is possible to achieve extremely high integration densities. The molecular nature of organic/polymer semiconductors allows the size of such transistors to shrink to such small dimensions, and also enables effective isolation between individual transistors. The dimensions of some of the interconnections, e.g. power and ground interconnections, may be significantly larger than those that appear in the figures.

[0044] The illustrative example given above and described in conjunction with FIGS. 1-17 is for a device with two levels of interconnect. The invention is equally applicable to ICs with one or even three levels of interconnect. The generic feature of the invention is the formation of at least one interconnect level prior to the formation of a field effect transistor over the interconnect level, with at least some of the sources or drains or gates of the transistors interconnected via the interconnect level.

[0045] Preferred embodiments of TFT devices according to the invention are made on flexible printed circuit substrates. Flexible printed circuit substrates are well developed and widely used. They are economical and can be used in hostile mechanical environments. They are also process and material compatible with polymer semiconductors. The TFT structure of the invention is shown with a flexible printed circuit substrate in FIG. 19. A portion of a flexible printed circuit substrate is shown at 51. The material of the substrate may be any suitable polymer such as polyimide. A commercial polyimide product, Kapton®, is available from Dupont. A typical thickness for the substrate is 30-100 microns. The substrate is provided with through holes which are plated to form plated-through-hole-interconnections from one side of the substrate to the other. A through hole interconnection is shown generally at 53 in FIG. 19. As is standard and well known in the flex circuit art, the through hole interconnection forms an electrical via for interconnecting portions of the printed circuit 54 on the bottom of substrate 51 with portions of a printed circuit 56 on the top side of the substrate. In the circuit illustrated here, the top interconnection level 56 is the gate level circuit, and via 53 is connected to MOS gate 57. The MOS gate dielectric layer is shown at 58, and the source and drain electrodes are shown at 61 and 62 respectively. The gate electrode 57 is shown thinner than the gate level printed circuit 56 to minimize the step for the gate dielectric and reduce the possibility of shorts or leakage in the MOS area. As mentioned earlier, the figures are not drawn to scale, and are schematic to illustrate the features of the invention. Thus in commercial embodiments, and as known to those skilled in the art, the thickness of the dielectric in the gate region may be thinner than layer 58 where it covers the printed circuit 56. Also, it is preferred that the source and drain electrodes not overlie the gate more than necessary to complete the inversion between source and drain, so as to minimize Miller capacitance. Accordingly, the actual gate electrode usually extends into the page of the figure as suggested by the gate illustrated in FIG. 20.

[0046] The processing required for fabricating the device of FIG. 19 is conventional. The flexible printed circuit substrate is typically drilled either mechanically or with a laser, or is mechanically punched with a suitable die, to form the vias. Typical vias are 150-300 microns in diameter. The metals forming the printed circuits may be copper, or copper with a gold strike. As known in the art they can be formed by either additive or subtractive processing, and the term printed circuit, as used herein, is intended to cover either of these alternatives. Typical thickness of the printed circuits is 1-5 μm Cu, optionally plated with 0.05-0.2 μm Au. The gate electrode itself is preferably much thinner, e.g. 0.1-0.5 μm, In some cases, and to simplify processing, the thickness of the gate electrode may be used for the entire gate level circuit. The gate dielectric layer 58 is preferably a polymer, such as polyimide, applied by spin coating. If more than one top metallization level is used, and interlevel vias are formed to provide interconnections, the interlevel dielectric may advantageously be a photodefinable polymer. The thickness of the polymer layer 58 may be in the range 0.05-0.2 μm. Layer 58 may also have dual thickness, a relatively thin layer over the gate and a thicker layer covering the printed circuit portions. Alternatively, the same objective may be reached by using different dielectric materials. A high k dielectric, i.e. a material with a dielectric constant greater than 4, may be used selectively over the gate. The active semiconductor layer for the TFT is shown at 65 and may be formed as described earlier.

[0047] A TFT device with a more complex flexible printed circuit interconnection arrangement is shown in FIG. 20. Here, three levels of metal interconnection are used on the top of the flexible printed circuit substrate. Optionally three interconnection levels may be used on the bottom surface of the flexible printed circuit substrate, or on both surfaces. The flexible printed circuit substrate is shown at 71, with top metal interconnection level represented by conductors 72, and bottom interconnection level represented by conductors 73. A through hole interconnection is shown at 74. The gate interconnection level is represented by 75, with interlevel via 76 interconnecting the first level and the gate level. The third interconnection level is represented by source electrode 77. The interlevel dielectric 78 is shown relatively thick, and the gate dielectric 79, between the gate level and the third level is shown relatively thin for reasons described above.

[0048] Various additional modifications of this invention will occur to those skilled in the art. All deviations from the specific teachings of this specification that basically rely on the principles and their equivalents through which the art has been advanced are properly considered within the scope of the invention as described and claimed.