In semiconductor technology, doping or doping (from Latin dotare , “equip”) describes the introduction of foreign atoms into a layer or into the base material of an integrated circuit . The amount introduced during this process is very small compared to the carrier material (between 0.1 and 100 ppm ). The foreign atoms are defects in the semiconductor material and specifically change the properties of the starting material, i.e. H. the behavior of the electrons and thus the electrical conductivity . Even a slight density of foreign atoms can cause a very large change in electrical conductivity.
If the electrical conductivity of semiconductors is to be changed, a distinction is made between p- and n-doping. With p-doping , foreign atoms are implanted, which serve as electron acceptors . With n-doping , however, electron donors are implanted. For changing the electrical conductivity of common semiconductor components made of silicon or germanium (the fourth main group ), the elements from the third main group such as boron , indium , aluminum or gallium come for p-regions and the elements from the fifth main group such as for n-regions for example phosphorus , arsenic or antimony are used.
Another application that is frequently used in microelectronics is doping silicon dioxide with boron or phosphorus. The resulting borophosphosilicate glass (BPSG) has a melting point 600 to 700 Kelvin lower than silicon dioxide. This makes the BPSG suitable, for example, for planarizing the wafer surface using a reflow process.
Using the example of silicon, the most widely used base material for semiconductor components, what is meant by n- and p-doping ( negative and positive doping ) will be briefly described below .
A silicon single crystal consists of tetravalent silicon atoms. The four valence electrons (outer electrons) of each silicon atom create four covalent bonds to its neighboring atoms and thereby form the crystal structure; this makes all four electrons binding electrons .
With n-doping (n for the freely moving negative charge that is introduced) pentavalent elements, the so-called donors , are introduced into the silicon lattice and replace tetravalent silicon atoms. A pentavalent element has five outer electrons available for covalent bonds, so that when a silicon atom is exchanged for a foreign atom in the crystal, an outer electron of the donor is (quasi) freely available (actually bound in an energy level just below the conduction band). The electron moves when a voltage is applied; this movement represents a current. At the point of the donor atom, a fixed positive charge arises, which is opposed to a negative charge of the freely moving electron.
With p-doping (p for the freely moving positive gap, also called hole or defect electron , which is introduced as a result) trivalent elements, the so-called acceptors , are introduced into the silicon lattice and replace tetravalent silicon atoms. A trivalent element has three outer electrons available for covalent bonds. For the fourth, an external electron is missing in the silicon crystal. This electron gap is known as a “hole” or defect electron. When a voltage is applied, this hole behaves like a freely moving positive charge carrier (in the valence band), it moves - analogous to the negatively charged electron - this movement represents a current. An electron - driven by the external field - jumps out of one covalent bond, fills a hole and leaves a new hole. In the place of the acceptor atom, a stationary negative charge arises, which is opposed to a positive charge of the freely movable hole.
The direction of movement of the holes is opposite to the direction of movement of the electrons and thus in the direction of the technical current direction .
A more detailed description of the electrical effects is provided by the ribbon model .
In electronics, one needs doping with different degrees of doping. A distinction is made here between heavy doping (n + ; p + ), medium doping (n; p) and weak doping (n - , p - )
|symbol||Conditions in Si||Ratios in GaAs
|n||1 donor / 10 7 atoms|
|p||1 acceptor / 10 6 atoms|
|n +||1 donor / 10 4 atoms||1 donor / 10 4 atoms|
|p +||1 acceptor / 10 4 atoms|
|p ++||1 acceptor / 10 3 atoms|
pn junction, components
By spatially adjacent different doping regions in the semiconductor, a pn junction with a space charge zone can be formed, for example, which has a rectifying effect in conventional diodes . Complex arrangements of several pn junctions enable complex components such as bipolar transistors to be formed in npn or pnp design. The terms npn or pnp in the case of bipolar transistors denote the sequence of the different doping layers. Thyristors or triacs , among other things, are formed with four or more doping layers .
Similar to inorganic semiconductor crystals, the electrical properties of electrically conductive polymers such as polyaniline (PANI) and organic semiconductors can also be changed by doping. By substituting carbon atoms in the chain structure of the polymer, the bond lengths change. In this way, intermediate energy levels arise in the energy bands of the molecule or the semiconductor as a whole, so-called polarons or bipolarons . Similar to inorganic semiconductors, doping is divided into two groups: oxidation reaction (p-doping) and reduction reaction (n-doping).
In contrast to inorganic semiconductors, the doping concentration in organic semiconductors can be in the percentage range. Such a high level of doping changes not only the electrical properties, but also all other properties of the material.
For the doping of semiconductors, there are basically four methods or techniques for introducing foreign atoms into a material:
- Alloy ,
- Diffusion ,
- Ion implantation and
- Neutron transmutation doping , d. H. Doping through nuclear transformation.
In the manufacture of semiconductor products, these techniques can be used alternatively or in addition to one another, depending on the application. The differently doped areas of bipolar transistors can be produced by diffusion, alloying or ion implantation. The choice of the appropriate technology depends on various requirements and framework conditions, e.g. B. process control, thermal budget in the overall process, available systems, contamination reduction or simply the costs.
Alloy technology is the oldest method for doping semiconductors in semiconductor technology. It is based on the controlled partial dissolution of the semiconductor through the formation of a surface metal-semiconductor melt and subsequent recrystallization.
In a first step, the dopant source is applied to the target material, for example by physical vapor deposition (PVD). The temperature is then increased, while part of the dopant diffuses superficially into the semiconductor and, for example in the case of aluminum, initially forms a metal silicide in silicon . This was followed by a further increase in temperature until the surface (the silicide) begins to melt. At the same time, further dopant diffuses into the semiconductor and these areas are also melted. The melting depth in the semiconductor is determined by the amount of deposited dopant and the solubility at the maximum temperature. This depends on the material combination and can be determined from the phase diagram . The amount of the deposited dopant therefore determines the alloy depth and thus the later position of the resulting pn junction. In the last step, the melt is slowly cooled so that it recrystallizes epitaxially on the semiconductor as a highly doped layer . The doping concentration shifts according to the solubility curve in the phase diagram. If the doping is not to take place on the entire semiconductor, the diffusion of the dopant and the formation of the melt in the corresponding areas must be prevented (locally). In the aluminum-silicon material system, this can be achieved with a sufficiently thick silicon dioxide layer, for example by thermal oxidation of silicon, photolithographic structuring and subsequent etching of the oxide layer.
Alloy doping is strongly influenced by the phase diagram of the material system. This means, on the one hand, not any dopants can be introduced into a semiconductor, and on the other hand, the doping concentration and also the position of the pn junction are severely limited. The best-known alloy systems are indium doping in a germanium crystal and aluminum doping in silicon. Nowadays, the method is no longer used in the volume production of semiconductor components. In addition to process challenges (e.g. cracks in the pn junction easily occur due to the brittleness of silicon alloys), it is also difficult to apply to the CMOS circuits that are common today .
Doping by diffusion
Diffusion is generally understood to mean a thermally activated equalization process for a concentration difference in a solid, in liquids or gases without any external influence (e.g. an electric field). If there is a difference in concentration, foreign atoms can penetrate another solid body at sufficiently high temperatures and move there. This can be done in three ways:
- Vacancy diffusion, d. H. through empty spaces in the crystal lattice
- Interstitial diffusion, d. H. between the atoms in the crystal lattice
- Change of place, d. H. Exchange of the lattice sites of neighboring atoms.
Diffusion processes in solids are described using Fick's laws . They depend on various factors:
- Material of the foreign matter and the target and their properties, e.g. B. crystal orientation,
- Concentration difference,
- Temperature as well
- Concentration of other dopants in the crystal.
According to Fick, how fast a dopant moves in the crystal is described by the diffusion coefficient of a substance. This depends on the size of the atom and the type of diffusion in the substrate; for example, the diffusion coefficient in silicon generally increases from arsenic to phosphorus to boron. Due to the small diffusion coefficient and the resulting necessary process time, arsenic is therefore practically unsuitable for introducing doping deep into the crystal, for example for producing the n-doped well of the CMOS process.
As mentioned, an important aspect for the diffusion and the resulting doping profile is the difference in concentration. Differences result primarily from the characteristics of the dopant source, therefore a distinction is made between two cases: 1.) Diffusion from an inexhaustible source and 2.) Diffusion from an exhaustible source. In the case of an inexhaustible source of dopants, it is assumed that the dopant concentration on the surface of the crystal is constant and that foreign atoms diffused into the depth are therefore replaced directly from the dopant source. This means that with increasing diffusion time and temperature, the dopant diffuses deeper into the crystal and the amount increases. The concentration on the surface remains constant. In practice, diffusion from the gas phase with the dopant concentration kept constant in the gas space can be viewed as an inexhaustible source of dopants. In the case of diffusion from an exhaustible dopant source, the amount of dopant is constant. With increasing diffusion time and temperature, the penetration depth of the dopant increases, but at the same time the concentration on the surface decreases. A practical example is the diffusion from a layer on the surface or the diffusion of the dopants after introduction by means of ion implantation.
As a dopant source, pure elements are usually not used because their vapor pressure is too low and they are difficult to vaporize. Therefore, light molecules that are generated from gaseous, liquid or solid sources are usually used. In semiconductor technology, typical gas sources for doping silicon are phosphine (PH 3 ), diborane (B 2 H 6 ) and arsine (AsH 3 ) in a carrier gas (argon, nitrogen) in a quartz furnace at temperatures of 800-1200 ° C is passed over the wafer. Typical liquid dopant sources are boron bromide (BBr 3 ) or phosphoryl chloride (POCl 3 ). They are brought into the carrier gas via a bubbler system and then passed over the wafer like the gas sources. The concentration in the gas space can be controlled comparatively easily via the bubbler temperature and the systems are easier and safer to handle. Solid diffusion sources are, for example, boron nitride or SiP 2 O 7 , which are placed as a “source wafer” or as a layer on a wafer between the wafers in the furnace. At high temperatures, some of this material diffuses into the gas space of the furnace.
In order to protect crystal areas from diffusion, the areas to be protected silicon dioxide are masked, that is, an approx. 300 nm thick silicon oxide layer is grown (see thermal oxidation of silicon ) and then locally removed in the areas for diffusion. Since the diffusion coefficient for typical dopants in silicon oxide is usually several orders of magnitude smaller than for silicon, the dopants cannot penetrate the oxide and thus cannot dop the silicon.
The natural, irregularly thick silicon also hinders diffusion from the gas space. In order to achieve a uniform introduction, a uniform thin thermal oxide is therefore often grown before the diffusion. In addition, a two-stage process is often used in practice, in which a certain amount of dopant is first brought into or onto the wafer at medium temperatures and then driven into the wafer at higher temperatures. In this way, depth of penetration and concentration can be better controlled. The first step can also be carried out as an oxidation in which, in addition to the dopant and the carrier gas, oxygen is also supplied to the gas space. A heavily doped silicate glass layer is formed which serves as an exhaustible source during driving, especially with phosphorus doping.
With ion implantation, charged (foreign) atoms ( ions ) are accelerated with the aid of an electric field and then directed onto the target (e.g. a silicon wafer). Corresponding production systems are referred to as ion implanters in semiconductor technology . The ions directed onto the target material penetrate the same and interact with it. Both elastic and inelastic collisions occur with the electrons and atomic nuclei. On the one hand, the ions are scattered, i.e. In other words, they experience a change in the direction of movement, and on the other hand they lose kinetic energy (inelastic collisions, electronic braking).
In contrast to diffusion, in ion implantation the maximum doping is not on the surface of the target material, but in the depth. The distribution of the ions in the target material depends on the properties of the ion (atomic mass, braking cross-section, energy, etc.) and the target material (atomic mass, density, crystal structure, crystal direction, etc.). Basically, it can be said that lighter ions (e.g. boron) can be implanted deeper than heavier ions (e.g. arsenic). The doping atoms are approximately normally distributed in the depth, so that a parabolic profile of the dopant concentration results in a semi-logarithmic representation. The mean depth of the ions is called the projected range . For a defined system, for example phosphorus ions on a silicon target, it is mainly dependent on the acceleration voltage for the ions and thus their kinetic energy. Strictly speaking, this only applies to amorphous targets (see LSS theory ). With crystalline and, above all, single-crystalline targets, the atoms are not evenly distributed on average and the crystal structure and its alignment with the ion beam can lead to major changes in the mean and individual penetration depth (see grid guide effect ). This can be reduced by a different angle of incidence of the ions, a slightly offset cut of the target material surface to the problematic crystal direction or by scattering layers. The level of the dopant concentration profile, on the other hand, is also determined by the “dose”, i.e. the number of ions per area. In semiconductor technology, the usual doses are in the range of 10 12 −10 15 cm −2 . The associated maximum dopant concentration (in cm −3 ) is usually 3–4 orders of magnitude higher. Since the ion implantation is not a non-equilibrium process, the doping concentrations can be adjusted well and can even be above the thermodynamic solubility limit.
The injected ions cause damage in the crystal due to the elastic collisions with the atomic nuclei, even with small doses. This breaks bonds in the crystal and moves the target's atoms. At very high doses, this can lead to amorphization on the surface, especially in the case of heavy ions, which is sometimes also brought about in a targeted manner. After the injected ions have given up their kinetic energy, they usually accumulate in interstitial spaces. These are not electrically active. In order to eliminate the crystal damage and to allow the implanted ions to diffuse onto electrically effective (active) lattice sites, the target is therefore subjected to a temperature treatment (the so-called “healing”). This usually takes place in an oven (e.g. oxidation oven) for 10–90 minutes at at least 900 ° C or, in order to keep diffusion low or better control it, by means of rapid thermal annealing (RTA) for a few seconds at similar temperatures . This is possible because it is primarily the maximum temperature and not the duration of the temperature treatment that is relevant. With the installation in the crystal lattice of the target, the doping profiles are also stabilized, since the diffusion on interstitial sites would take place much faster. It should be noted, however, that not all ions are incorporated into the crystal lattice during the temperature treatment, depending on the method used and the process parameters.
When manufacturing semiconductor components by means of ion implantation, as with diffusion, all points should never be doped equally. Areas that should not be doped are covered with a mask. A photoresist mask of the appropriate thickness is often sufficient. So-called hard masks made of silicon oxide, silicon nitride or polysilicon are also used. In addition, existing structures on the target can be used for self-aligning ion implantation processes (see spacer technology ).
Nuclear transmutation initiated by neutron doping
With some materials, e.g. B. silicon, doping can also be achieved by irradiation with neutrons, for example in a heavy water reactor. The mass number of some atoms is increased by one through neutron attachment. This can lead to stable as well as unstable nuclei which, depending on their half-life , transform an isotope of another element, for example through beta decay. In special cases, foreign atoms can be "introduced" into a target, for example low phosphorus doping (<10 14 cm −3 ) with the highest uniformity in a larger volume of silicon. Higher doping is possible, but is associated with significantly longer process times (> 100 h) and is therefore practically irrelevant.
Advantages of neutron transmutation doping compared to doping during crystal pulling are the much higher uniformity without the formation of striations , that is, doping or interference fluctuations in the single crystal. The method was originally used for the substrate production or doping of power semiconductor components, which are very sensitive to Disturbances and fluctuations are. However, it can be seen that the technical effort is also problematic for reasons of radiation protection, which is why the single crystal manufacturers have been working for several years to improve the doping during crystal growth and have made progress here.
Comparison of the doping methods
Diffusion is characterized by the following properties:
- flat concentration profiles and therefore wide pn junctions,
- The dopant concentration decreases with depth and is determined by the process time and temperature.
- strong interaction several doping processes, e.g. B. Deterioration of the profiles and outdiffusion
- exponential temperature dependence poorer reproducibility
- Lateral spreading of the dopants leads to a large area requirement and very high integration densities are no longer compatible
- simple process in which many wafers can be processed in one furnace at the same time and is therefore inexpensive
The ion implantation, however, is characterized by the following properties:
- steep pn junctions due to high reproducibility and low thermal load
- Implantation independent of the solubility limit
- Manufacture of "buried" areas
- Manufacture of areas with similar doping concentration through coordinated multiple implantation
- Inexpensive local implantation through the use of photoresist masks
- wide range of implantable elements
- high technical effort (acceleration, high vacuum, etc.) and therefore comparatively expensive
- very limited doping of structures with high aspect ratios, e.g. B. deep trenches.
In microsystem technology, too, areas or layers are specifically doped in order to change the material properties. The aim here is less to influence the electrical properties than to change chemical or mechanical properties. A typical example is the creation of an etch stop layer for the (anisotropic) wet chemical etching of silicon, through the targeted boron doping of silicon with concentrations greater than 5 · 10 19 cm −3 . Such high boron doping leads to a high concentration ν of defect electrons which recombine with the silicon electrons. For this reason, only a few electrons are available for the redox reaction of alkaline etchants with silicon, which leads to a decrease in the reaction rate and thus the etching rate. In return, the installation of large amounts of boron leads to a change in the spacing of the grids and thus to mechanical stress. This must be taken into account if the etch stop layer is to be used later, for example as a membrane in a pressure sensor.
- Hilleringmann, Ulrich: Silicon semiconductor technology: Fundamentals of microelectronic integration technology . 5th, erg. U. exp. Edition. Vieweg + Teubner, Wiesbaden 2008, ISBN 978-3-8351-0245-3 , pp. 95 ff .
- Ulrich Hilleringmann: Silicon semiconductor technology . 4th, through and additional edition. Teubner, Stuttgart / Leipzig / Wiesbaden 2004, ISBN 3-519-30149-0 , pp. 105 ff .
- Dieter Sautter, Hans Weinerth, Sautter, Dieter: Neutron doping . In: Lexicon Electronics and Microelectronics . 2nd, updated and exp. Edition. VDI Verlag, Düsseldorf 1993, ISBN 3-18-401178-X , p. 721 ff .
- Josef Lutz: Semiconductor power components: physics, properties, reliability . 2nd Edition. Springer Vieweg, Berlin / Heidelberg 2012, ISBN 978-3-642-29796-0 , p. 55 ff .