7
Advanced silicon radiation detectors in the vacuum ultraviolet and the extreme ultraviolet spectral range
Abstract
In this chapter, the numerous challenges in reliable measurement of vacuum ultraviolet (VUV) and extreme ultraviolet (EUV) radiation with silicon radiation detectors will be discussed. The most severe issue is the high absorbance of any kind of material in these spectral ranges, which poses stringent requirements on possible detector designs. In particular, a high value of the spectral responsivity and the necessary radiation hardness put contradicting demands on the devices. To find a solution, it is necessary to describe and understand in detail the basic physical processes for the interaction between the radiation and the detector. The spectral responsivity of VUV and EUV detectors is measured using monochromatized synchrotron radiation as a radiation source and electrical substitution radiometers as primary detector standards. We will present selected types of silicon-based semiconductor detectors and discuss their performance in the context of the named aspects.
Keywords
Radiation hardness; Silicon photodetectors; Spectral responsivity; VUV/EUV radiation; VUV/EUV radiation damage
7.1. Introductory overview
Radiation detection in the vacuum ultraviolet (VUV) and extreme ultraviolet (EUV) spectral ranges comes with specific physical and technical challenges, which are not present in, e.g., the visible (VIS) and infrared (IR) regimes. However, this part of the electromagnetic spectrum with wavelengths shorter than the air UV is still of growing interest for industrial applications. The semiconductor photolithographic industry with its process lines at wavelengths of 193 and 13.5 nm requires reliable radiant power measurements and may be seen as a main driving force. Other applications (e.g., UV curing for varnish hardening or water disinfection) also have a certain presence in this field. Moreover, analytical methods used in process control, such as ellipsometry for the determination of thin film thicknesses, are being extended to short wavelengths as well and also need reliable VUV and EUV detectors.
In the following, we denote the spectral range from 40 to 200 nm (the working range of normal incidence reflective optics) as VUV and the adjacent shorter wavelength range from 40 nm down to about 1 nm (where grazing incidence mirrors—or, for restricted wavelength intervals, normal incidence multilayer mirrors—are used) as EUV. In both spectral ranges, due to absorption, the radiation must be transported through vacuum. Moreover, particularly in the VUV range, the radiation is almost completely absorbed within a layer of only a few nanometers thickness of any solid material. The radiation is sufficiently energetic to ionize through the photoelectric effect and therefore charges insulating materials. This puts stringent requirements on any radiation detecting device. Particularly, the sensitive region of the detector must be as close to the upper surface as possible, which means that there should be no protecting top layer. Silicon-based radiation detectors are the common choice for these wavelength ranges because they represent a very mature technology on a highly sophisticated level. In the following section, we will discuss the basic device solutions to meet the requirements of VUV and EUV detection and present the respective methods for radiometric characterization of the detectors.
7.2. Challenges for radiation detection in the VUV and EUV spectral ranges
In the context of this chapter, we will restrict the discussion to the detection of photons by the photoelectric effect. This means that we will not consider any other excitation processes (e.g., as in photochemical detectors) or purely thermal detection principles. Moreover, we will neither discuss single-photon detectors nor energy dispersive detectors, although most of the considerations presented here will apply to these detectors, too. The photoelectric absorption creates at least one excited electron in an energy state above the vacuum level. If the absorption occurs close enough to the surface, this electron, as well as secondary electrons generated by electron-scattering processes, may escape from the detector to the vacuum, generating an external photocurrent. If the electrons do not escape into vacuum, they populate the conduction band in the detector material. Thus, a first classification of the detectors can be based on whether the detected electrons are external or internal.
The essential part of external photoemissive detectors is the photocathode, where electrons are ejected by photon excitation. In photoemissive diodes, these electrons are directly measured as a current, whereas in photomultiplier tubes, secondary electrodes (dynodes) are used for a charge multiplication process of the primarily generated electrons. These detectors may be operated in a proportional amplification mode or in a saturated photon counting mode. A well-known example for a photoemissive diode is the so-called “gold diode” where just the external photocurrent from an Au-coated electrode is measured. From the known photoelectric yield of Au in the VUV spectral range (Henneken et al., 2000), the measured current can be directly related to the radiant power. In metal vacuum chambers, it is principally sufficient to place only the photocathode metal plate insulated from ground into the photon beam and apply a negative bias voltage. The surrounding vacuum chamber, which is usually grounded via other installations such as ion pumps and vacuum gauges, acts as the anode. The accuracy of the measurement is limited because the photoelectric yield is already strongly influenced by submonolayers of contamination. However, the method is simple and robust, and the results will at least be correct in the order of magnitude, which is not necessarily assured for detectors based on internal photocurrent measurements if they were subject to prolonged irradiation. Addition of an external anode results in a more sophisticated design and assures collection of the emitted electrons. However, it does not remove the main source of error in the measurement, which is the cleanliness of the cathode surface. In particular for the VUV range cathode materials with low work function (usually containing alkali metals) are in use, however, their surfaces—and thus, their photoelectric yield—are even more sensitive to changes. Therefore, such cathodes are preferably used in sealed tubes behind some entrance window (magnesium fluoride) what limits the spectral range to wavelengths longer than 120 nm and, moreover, shifts contamination issues now to the window surface.
On the other hand, detectors using the internal photoeffect offer higher responsivity and potentially higher accuracy, if operated properly. These detectors measure the change in their electrical properties induced by the photogenerated charge. The internal photoeffect excites an electron from an occupied band, inner shell, or valence band into a state above the Fermi level. To induce any difference in the electronic properties of the detector material, the band into which the electrons are excited must be almost unoccupied such that the electron population induced by the photon irradiation significantly increases the initial population. Ideally, there are no free charge carriers in this band without photon irradiation. Thus, the internal photoeffect leads either to increased conductivity in a homogeneous high-resistance material or to the creation of electron–hole pairs of charge carriers, which are then separated by an internal electrical field in a depleted region. Furthermore, the charge carriers created inside the material must be measured at external electrodes. Therefore, not only are materials with a sufficiently long carrier lifetime and adequate carrier mobility needed but also some kind of technology for creating contacts to the detector material.
For the photoconductive type of detectors, only one type of material of sufficiently high resistivity of either n- or p-type is needed; no doping technology to create both n- and p-conducting regions to form a p-n junction at the p-to-n interface is required. Therefore, the first examples of detectors from more exotic materials such as diamond are, indeed, of the photoconductive type (BenMoussa et al., 2003). The use of such wide bandgap materials is mainly triggered by the request to detect very low levels of VUV radiation in the presence of very high levels of VIS and IR radiation (e.g., in space-based solar physics applications) (Schühle and Hochedez, 2013). These photoconductive detectors, however, do have many drawbacks: an unstable temporal response, limitations in size, homogeneity issues caused by the indispensable contact grids, and a prolonged response time not suited to measure individual pulses of high-frequency pulsed radiation sources.
Any kind of detector used in the VUV and EUV wavelength ranges must be in a vacuum because the radiation to be detected will be strongly absorbed by any atmosphere. Furthermore, the radiation absorption of all materials in the spectral region around 100 nm is almost equally high, resulting in full absorption of the radiation within the first few nanometers (Fig. 7.1).
No solid material is transparent for wavelengths shorter than 105 nm, which is the absorption-edge wavelength of LiF, corresponding to its bandgap energy of 11.8 eV. This particularly means that no window or other transparent layer can be used in front of the detector for protection. For radiation detection with photodiodes, this additionally implies that the sensitive region of the detector must be as close to the upper surface as possible, which severely limits the use of protecting top layers. Another specific issue for the detection of VUV (and shorter wavelength) radiation is that the photon energy is sufficiently high to create photoelectrons (and respective vacancies) by external photoemission. This is particularly the case in any insulating top layer such as silicon dioxide (SiO2), which is widely used as the natural passivation layer on silicon. In the UV and VIS spectral ranges, SiO2 is a very appropriate option for a top layer because it is transparent and can, additionally, be used as an antireflection layer. However, vacancies resulting from photo-induced electron emission due to VUV and EUV irradiation will not be quickly refilled and, as a result, a permanent positive space charge is generated. The resulting electrical field superposes the internal space charge of the p-n junction. For p-on-n diodes, this lowers the junction potential and may even inverse it at the oxide interface so that detector responsivity drops to zero.
Figure 7.1 Half-value layer thickness of representative materials in the extreme ultraviolet (EUV) and vacuum ultraviolet (VUV) wavelength ranges. Examples for UV-transparent, insulating materials are SiO2 (– – –) and MgF2 (----). Si (–––) is shown as the typical active detector material and Pt (.......) as a representative for a front metal contact layer. The spectral ranges of EUV and VUV as denoted here are also indicated, as well as the most important wavelengths for semiconductor lithography in the respective spectral range. The visible (VIS) spectral range is indicated for reference. The data are taken from literature for the EUV spectral range (Henke et al., 1993) and for the VUV range (Palik, 1985).
The external photoemission generates a current from the top surface of the diode through the vacuum toward ground. Therefore, two components of the total current must be considered: an external current through vacuum to ground and an internal current toward the reverse of the diode (see Fig. 7.2). For pixel sensors or integrated sensors with several diodes on the same chip, usually the reverse of the diode is used as a common electrode and the contacts on the front are used to read the signal from each individual pixel of the detector. For the VUV spectral range, this approach results in the measurement of the sum of external and internal photocurrent. Any instability or change of the external photocurrent (e.g., due to surface contamination and local charge buildup) will thus be included in the measured signal.
Finally, the most challenging issue for VUV and EUV photodetectors comes from radiation-induced degradation. Three main processes must be considered: (1) charging of the surface layers, (2) the buildup of contaminating adsorbates from the residual gas atmosphere, and (3) damage of the surface passivation by prolonged irradiation. The combination of the relatively high photon energies sufficient for the creation of photoelectrons, the extremely short absorption lengths, and the need of operation in vacuum results in strong irradiation aging of detectors, which otherwise are regarded as quite stable in other (optical) wavelengths ranges.
Figure 7.2 Scheme of a photodiode with external and internal photocurrent. For p-on-n diodes, both components have the same polarity; for n-on-p diodes the polarity is opposite. The charge separation in the depletion region is shown in the figure for an n-on-p diode. Only if the contact on the front face is put to ground and the current is measured from the bulk side electrode, the internal photocurrent alone is measured.
To summarize, for the detectors using the internal photoeffect, the external photoemission combined with the strong absorption of radiation in the VUV and EUV spectral ranges causes a significant operational challenge that is not present in applications in the VIS and near UV spectral ranges. To meet these challenges, feasible detectors must have a design adapted to the critical issues presented here.
7.3. Device solutions for radiation detection in the VUV and EUV spectral ranges
The demanding challenges for a stable and highly responsive VUV and EUV radiation detector require devices with a highly uniform depletion zone and shallow cross-section. Therefore, any solution for these challenges in the fabrication of a device must be based on a versatile and mature technology. This strongly favors silicon as the detector material. Silicon technology outperforms any alternative technology—at least, if the integration of further functionality is required to make the sensors smart. For some applications of VUV detectors, however (e.g., solar physics), it is desirable to have a strong suppression of the near UV and VIS spectral ranges because the radiant power in these spectral ranges is much higher than that in the VUV. This requirement led to the development of wide bandgap material detectors (known as solar-blind detectors) (Schühle and Hochedez, 2013) based on gallium arsenide (GaAs), aluminum nitride (AlN), aluminum gallium nitride (AlGaN), silicon carbide (SiC), or even diamond (BenMoussa et al., 2006). Based on these materials, mainly photoconductive, Schottky-type (Malinowski et al., 2011), and metal-semiconductor-metal (Brendel et al., 2015) devices have been realized so far. Although single devices have demonstrated remarkable results (e.g., spectral responsivities comparable to those of Si photodiodes at certain wavelengths) (Kalinina et al., 2016) the current technologies for these detector materials have not reached the same degree of sophistication as in the case for Si. At the moment, this limits the use of those detectors mainly to astrophysical applications (Schühle and Hochedez, 2013). For the EUV spectral range, metal filters on silicon provide an alternative solution because metals such as Zr or Al are somewhat transparent in the wavelength region below 40 nm but, being metals, are opaque to UV and VIS radiation (Canfield et al., 1994; Seely et al., 1999). As outlined in the previous section, VUV and EUV radiation interacts with the detector by a process of photoelectric absorption close to the surface. For some applications in the field of hard X-ray or IR detection, it may be desirable to hybridize a highly absorbant material such as cadmium telluride (CdTe) with silicon for use in the signal readout and amplification (Rügheimer et al., 2008). In the VUV and EUV ranges, however, silicon itself is sufficiently absorbant. Therefore, silicon is the first choice for VUV and EUV detectors because it provides low-cost solutions to achieve superior homogeneity, very good electrical properties, and long-term stability, as will be discussed in Section 7.5.
For the VIS and near UV spectral ranges, the most common diode structure is a silicon p-on-n junction protected by a SiO2 passivation layer, which is the natural passivation top layer and also acts as an antireflection layer. Because the SiO2 half-value layer thickness (i.e., the thickness at which half of the photons entering the layer are already absorbed by the material) is only 4 nm in the VUV spectral range (Fig. 7.1), an obvious step toward VUV-sensitive diodes was to thin the oxide down to a few nanometers. Initial investigations on the VUV detection properties of silicon diodes were reported in the 1980s (Korde and Geist, 1987). The authors compared inversion layer diodes (Geist et al., 1981), n-on-p diodes, with P and As doping (Korde and Geist, 1987) as well as B-doped p-on-n diodes. Besides issues related to impurity—which were typical issues for diode technology at this time—the surface charging of the oxide was identified as the main issue for device stability. With this consideration in mind, n-on-p doping polarity was suggested because, in this case, the positive surface charge does not counteract the internal field of the junction. Further improvements could be achieved by nitridation of the silicon-oxide interface (Korde et al., 1993). Nevertheless, the stability of these devices is not sufficient for higher radiant power applications. For exposures above about 1 J/cm (BenMoussa et al., 2003), these diodes still degrade (Scholze et al., 1996). As a result, devices without oxide passivation layer were developed. For the n-on-p type devices, the SiO2 layer was substituted with a metal-silicide layer (Korde et al., 2003), yielding devices with much higher stability.
In the VUV spectral range, it is crucial to achieve a monotonous and very steep doping profile to extend the build-in electrical field as close to the surface as possible to collect all generated charge—the photons are absorbed close to the surface because of the low penetration depth. Different doping processes had been used to achieve this, using elements such as P, As, or B as donors or acceptors (Korde and Geist, 1987). In the case of B-doped p-on-n diodes, the junction as produced by standard doping technologies is not sufficiently shallow. This has been studied in detail for the production of energy dispersive EUV detectors (Hartmann et al., 1996). Energy dispersive detectors measure single photons and convert the charge created by each single photon into a voltage pulse, which is proportional to the charge created. These pulses are registered with a digital pulse-height analyzer and incrementally stored in a respective channel to create a pulse-height spectrum, which represents the measured photon energies. Here, photons in the spectral range from 4 to 10 nm are detected individually. Because of the silicon L-absorption edge, the penetration depth of these photons in silicon is even lower than for the spectral range from 12.5 to 60 nm (see Fig. 7.1). Furthermore, the charge created per photon in the energy range above the Si-L edge is sufficient to detect charge losses by recombination for each individual photon in the measured pulse-height spectrum. By measuring the pulse-height spectrum, one can differentiate between a situation where every photon is registered but only a part of the initially generated charge is collected and a situation where the full charge is collected for each photon that is detected, but a proportion of the photons is lost by absorption in a dead layer on top of the detector. For the measurement of photocurrent, i.e., by averaging over a large number of photons, both situations would yield the same result. Therefore, the energy dispersive detection of the low penetration depth but rather high-energy photons with photon energies higher than 100 eV is a good means by which to understand the surface-related recombination losses for photons absorbed in the uppermost region of a silicon detector. The relation between the spectral line shape in energy dispersive photon counting and the charge collection efficiency as a function of depth close to the detector entrance contact is presented elsewhere (Hartmann et al., 1996; Scholze and Procop, 2009).
The basic conclusion from these investigations is as follows: In the EUV spectral region, each photon generates a number of electron–hole pairs, which is proportional to the photon energy. The proportionality factor is 1/W, where W is the electron–hole pair creation energy (Scholze et al., 1998). If there is a loss of charge because of recombination at the surface, the charge measured per pulse is reduced. Thus, the measured line shape for monoenergetic photons yields direct information on recombination-induced signal losses. By this method, it was shown that standard B-implanted junctions did not achieve sufficient field strength close to the surface to pull the electrons away from the surface sufficiently rapidly. For the energy dispersive detectors in this study, the build-in field strength was improved by an additional P-implantation just beneath the B to increase the local field strength (Hartmann et al., 1996) close to the surface.
Another approach is “pure boron” technology (i.e., a boron-doped p-on-n diode with no oxide passivation layer) (Nanver, 2011). Chemical vapor deposition of a thin amorphous boron layer on top of the diode solved the issue of implanted B doping because it provides a B reservoir at the surface and thus assures a stable and steep monotonous doping profile, which results in a high build-in electric field extending up to the surface, facilitating an efficient charge carrier collection, even close to the boron–silicon interface. A similar type of radiation hard and highly sensitive boron-doped diode with an SiO2-free top layer was independently developed elsewhere (Aruev et al., 2009). Finally, also PtSi/n-Si Schottky diodes proved to be extremely radiation hard in the VUV spectral range (Solt et al., 1996). However, their shunt resistance is relatively low, which challenges the measurement of a low photocurrent in the picoampere (pA) range. For all approaches, not only the design determines the detector's properties but also the stability of the manufacturing process because the characteristics strongly depend on the purity and uniformity of the layers and interfaces.
So far, only single-diode devices have been discussed. For many applications, however, either imaging sensors are required or several small diodes will be integrated on a single chip for advanced detection systems. This requires the readout of several separate signals from the same chip. The natural solution would be to use a common electrode on the reverse and create the requested detectors with individual electrodes on the front face. In this case, however, the signals must be read from the front. This is a common technique for pixelated detectors in the UV and VIS spectral ranges and is also offered for EUV detectors (optodiode). The problem is that, for this polarity, the external photocurrent is included in the measured signal. This might be acceptable for EUV detection where the internal current is much higher than the external, and thus a small change in the external current does not significantly change the total signal. In the VUV range, however, both current components are about equal. Therefore, another solution is needed. This can be achieved by adding a further p-n junction between the bulk side of the sensitive junction and the substrate of the detector. Then, the sensitive junction can also be individually contacted on its reverse, whereas the front surfaces of all diodes are connected to a common contact, which is put to ground (see Fig. 7.3). There is, however, a price to be paid for this configuration: the additional parasitic p-n junction in parallel to the photodiode will contribute to the dark current and the output capacitance. There are more sophisticated systems based on double-sided wafer processing and charge-drifting systems, such as that for active pixel detectors (Treis et al., 2005). In particular for applications in space-based instruments, intensifiers are being used, which convert VUV and EUV radiation into VIS light by means of a microchannel plate (MCP) and a phosphor screen (Schühle, 2013). Although this avoids the issues for VUV and EUV detection for the sensor, the MCP itself—on the other hand—significantly enlarges the imaging device and adds further complications such as the need to use a high-voltage supply and image distortions. Moreover, the MCP is subject to irradiation damage as well as the phosphor might be irreversibly damaged by high intensities.
7.4. Methods of radiometric investigation and characterization
In radiometry, the basic physical quantity to be determined is the radiant power of the electromagnetic radiation. Commonly, a radiation detector is described in terms of its spectral responsivity (i.e., its output signal in relation to the incoming radiant power). With the exception of ideal (totally “black”) absorbers, spectral responsivity is a function of the wavelength. This is particularly the case for silicon-based photodiodes, where the material characteristics (e.g., surface reflectance, internal quantum efficiency) show strong dependencies from the wavelength. In the common event that the radiation has a continuous-wave characteristic, the current of the silicon photodiode in the short circuit (photovoltaic) mode is usually taken as output. It is also possible to apply bias voltages if needed and, in the event of pulsed radiation, the detector output is then measured in terms of the total charge created, or the voltage pulse height at a certain load resistance—which, for a fixed temporal shape of the pulse, is proportional to the charge generated. The detector's response, however, cannot be assumed to be a unique constant; it may vary across the detectors surface (spatial nonuniformity) with the radiant power (saturation and other nonlinearities), temperature, and time (degradation by irradiation and long-term temporal instability). Moreover, especially if the measurement is taken under an oblique angle of incidence, the responsivity is polarization-dependent due to the surface reflectance.
Figure 7.3 Operational scheme of pixelated diode detectors with top electrode at ground potential. The individual pixels are insulated with respect to the substrate by an additional p-n junction. Note that the substrate has the same polarity as the entrance contact of the diodes in this case. The example shown is for a p-on-n sensitive junction formed in an n-conductive epi-layer on a p-conductive substrate.
Basically, any physical measurement must be traceable to the SI (International System of Units): the quantity to be measured must be related to the respective realization of the unit, the “primary standard.” In the case of radiant power, electrical substitution radiometers as primary detector standards had already been introduced with the establishment of quantitative radiometry at the end of the 19th century (Kurlbaum, 1894). Since the early 1980s, their performance was improved substantially by turning to cryogenic temperatures (Martin et al., 1985). Electrical substitution radiometers are thermal detectors based on the equivalence of radiant and electrical heating. By substitution of the absorbed (radiant) heating power to be measured with an electrical heating power, traceability of the measurement to basic electrical units is established. For the long wavelength range (i.e., in the (air-) UV, VIS, and IR spectral ranges), measurements are commonly realized using (tunable) lasers or monochromatized radiation from arc discharge sources. However, for wavelengths shorter than 200 nm, measurements in air are no longer possible because of the absorption of air-borne oxygen and water. Although it is possible to some extent to use nitrogen (or noble gases) for purging, at least the absorption edge of available window materials gives a definite limit at about 105 nm (see Section 7.2). To cover the whole VUV and EUV spectral range, the use of monochromatized synchrotron radiation for these purposes has been established by certain National Metrology Institutes (NMIs): the Physikalisch-Technische Bundesanstalt (PTB) in Germany (Gottwald et al., 2006), the National Institute of Standards and Technology in the United States (Arp et al., 2003), and the National Metrology Institute of Japan (Saito and Onuki, 1992). Although this is quite an effort from the perspective of instrumentation, the use of synchrotron radiation offers several advantages for radiometric measurements: besides covering a wide spectral range, the radiation characteristics are well known and the source itself is a high-vacuum machine, providing clean ultrahigh vacuum conditions for the measurements. The mutual agreement of the different existing standards of the NMIs mentioned above in their measurement results within their individual uncertainties was proven by an international validation process in recent years (Scholze et al., 2010; Gottwald et al., 2011; Tanaka et al., 2012).
At PTB, synchrotron radiation from the electron storage rings Metrology Light Source (MLS) (Klein et al., 2011) and BESSY II (Klein et al., 2002), both located in Berlin-Adlershof, is used at different beamlines either for the UV and VUV spectral range down to 40 nm (Gottwald et al., 2010) or for the EUV spectral range from 40 down to 1 nm, the latter covering the 13.5 nm wavelength with its relevance for EUV lithographic technology. For these two spectral subranges, two different cryogenic electrical substitution radiometers are also in use at PTB as primary detector standards (Rabus et al., 1997, 2002), ensuring the lowest uncertainty possible for the respective range. Both radiometers are equipped with cavity absorbers, which ensure almost complete absorption of the radiation in the respective wavelength ranges. Beyond the EUV range, at dedicated X-ray beamlines, it is also possible to extend detector calibration and characterization up to 60 keV photon energy, respectively, 0.02 nm (Fig. 7.4) (Gerlach et al., 2007). The existing instrumentation allows the detectors to be precisely calibrated with a low level of uncertainties and, furthermore, to be characterized in detail regarding their temporal characteristics, temperature dependence, spatial uniformity, linearity, and aging behavior.
7.5. Spectral responsivity and radiation hardness of VUV and EUV radiation detectors
The elementary physical process determining the spectral responsivity of a semiconductor photodetector is the creation of electron–hole pairs in the photodiode depletion zone. The mean energy required to create an electron–hole pair W is a material property of any semiconductor. In the UV and VIS spectral range (i.e., for wavelengths above 310 nm), it equals the photon energy: there is exactly one electron–hole pair created per absorbed photon—that is, the quantum efficiency ε is constant ε = 1 (electron/photon). At shorter wavelengths (or higher photon energies, respectively) the kinetic energy of the initial photoelectron is sufficiently high to generate secondary electron–hole pairs by electron–electron scattering. For silicon, the probability of these additional ionization processes, however, stays low up to about 4.4 eV photon energy, corresponding to a wavelength of 280 nm (Scholze et al., 2000). Thus, ε = 1 holds essentially down to this wavelength and the corresponding value for W increases to W = 4.4 eV. With increasing photon energies (i.e., decreasing wavelengths) the secondary ionization process sets in and W converges rapidly to a constant value. This asymptotic value is almost reached at 80 nm in the VUV spectral region. For shorter wavelengths (i.e., in the EUV spectral range) the value is constant. For silicon, it has been determined to be W = (3.66 ± 0.03) eV (Scholze et al., 1998, 2000).
Figure 7.4 Scale of spectral responsivity maintained by Physikalisch-Technische Bundesanstalt using monochromatized synchrotron radiation from UV to X-ray regimes as source and calibrated silicon photodiodes as detectors (traceable to cryogenic radiometers as primary detector standards). The data shown are for an n-on-p silicon diode with a thin SiO2 passivation layer. The symbols indicate the different beamlines used: double-crystal monochromator at a wavelength shifter (○), four-crystal monochromator at a bending magnet (◇), EUV radiometry beamline (△), plane grating monochromator at an undulator (●), and the normal-incidence monochromator at the Metrology Light Source (♦). The solid line is the model estimation for best-effort silicon radiation detectors. The severe drop in measured responsivity at wavelengths shorter than about 0.2 nm is simply due to the high-penetration depth of the X-rays and the only μm-thick sensitive depletion region of the diode.
For an ideal detector, the spectral responsivity, s, would only be characterized by this value, s = e/W, where e denotes the elementary charge (see Fig. 7.5). To obtain the spectral responsivity of a real detector, additional physical processes must be taken into account. In the EUV, the detector surface does not play any role as the reflectance of all materials goes to zero for wavelengths below about 50 nm. In the VUV spectral range, the reflectance of the detector surface must be taken into consideration. Between 100 and 200 nm, a silicon/silicon oxide surface has a reflectance higher than 50%. Note that, unlike in the UV and VIS spectral ranges, the selection of a λ/4 thickness of the oxide to achieve an antireflection behavior is not an option because of the strong absorption in the oxide of refractive index ñ = n – ik, with k being much larger than 1. The absorption in (dead) surface layers must be taken into account in all wavelength ranges, by pointing out that in the VUV this is the by far dominating term.
Figure 7.5 Responsivity of ideal silicon diodes. The solid line is the limit resulting from the internal quantum efficiency of silicon (Scholze et al., 2000). The straight-dotted horizontal line is the short wavelength approximation with constant W = 3.66 eV, and the steeply increasing dotted line above 200 nm represents the approximation for the long wavelength range with constant quantum yield ε = 1. The long dash-dotted line includes the loss by external reflection for a 3 nm thick SiO2 layer, calculated using the data from literature (Palik, 1985). This might by tailored in the region where the oxide is transparent by using thicker layers; the behavior shown is for a 75 nm thick oxide is indicated by the long dashed line. Below 130 nm even a thin oxide begins to absorb. The absorption loss for a 3 nm thick oxide layer is additionally included in the data and is shown by the short-dashed line (below 120 nm, where the oxide becomes transparent, it overlaps with the dash-dotted line accounting for reflection only).
The operation of any semiconductor device requires a passivation of surface electronic states to prevent excessive recombination, at least if the electronically active regions are close to the surface, which is a “must” for VUV detection. For silicon, this passivation is usually achieved by a thin layer of grown silicon oxide. For improved radiation hardness, however, nitrided oxide or metallic silicide layers are also in use. In the spectral region of highest absorption, around 100 nm, a 5–6 nm silicon oxide layer already absorbs about 80% of the incoming radiation. Further loss mechanisms are due to electron–hole recombination and escape processes, where electrons are lost because of leaving the depletion zone. Although these effects are unavoidable, their individual strength strongly depends on detector quality (e.g., the number of impurity traps or the uniformity of interface layers between depletion and passivation layers). In the VUV spectral range, the absorption length of any kind of material reaches its minimum of, typically, only a few nm, and therefore any change of the detector surface—even of just one monolayer—has a drastic effect. The strong absorption in the surface layers typically reduces the spectral responsivity at 190 nm by about a factor of 5, as compared with the value at 13.5 nm (see Fig. 7.4). Furthermore, the whole energy of the photons is deposited within a very thin layer at the surface of the detector while, at the same time, the detector is highly sensitive to any changes in these uppermost surface layers. This results in strongly enhanced radiation damage in the VUV range in comparison with, for example, the EUV spectral range.
From these considerations, the limiting performance for an ideal silicon photodiode in the VUV and EUV spectral ranges can be estimated, as shown in Fig. 7.5. For short wavelengths below about 10 nm, the intrinsic limit, s = e/W, can be almost realized with actual detectors because there are no reflectance losses and a thin passivation layer does not excessively absorb. Above 10 nm, however, the absorptance of even the thinnest passivation layers increases drastically to about 50% in the spectral region around 100 nm. Representative experimental results for the spectral responsivity of different types of silicon photodiodes are shown in Fig. 7.6. Note that the radiation-hard devices with metal-silicide passivation have significantly lower responsivity in the VUV spectral range due to the strong absorption of these rather thick passivation layers. They are also less efficient in the EUV spectral range below the silicon L-absorption edge at 12.4 nm. From spectral modeling of the device performance (see e.g., Scholze et al. 1998), it can be concluded that this effect is not due to a thin dead layer of silicon but must be explained by a charge collection efficiency monotonously increasing from a starting value below 1 to unity deeper in the active region. This reduced charge collection efficiency for photons absorbed close to the surface hints at charge losses at the silicon–silicide interface. Thus, it indicates the less efficient passivation behavior of the silicides as compared with SiO2. This effect also contributes to the lower efficiency of silicide-passivated diodes in the VUV spectral range. There, however, it is masked by the strong absorptance of the passivation layer itself.
Figure 7.6 Absolute spectral responsivity of different types of photodiodes in the vacuum ultraviolet (VUV) and extreme ultraviolet (EUV) spectral ranges. Shown are data representative of n-on-p diodes with a thin nitrided oxide passivation or a metal-silicide passivation (optodiode). Boron-doped p-on-n (Aruev et al., 2009) and “pure B” (Nanver, 2011) devices achieve responsivity in the range between n-on-p diodes with a thin layer of nitrided oxide and PtSi/n-Si Schottky devices (Solt et al., 1996). The theoretical estimation from Fig. 7.4 is also shown as a benchmark for comparison (----). There were no representative data available for the PtSi/n-Si Schottky devices in the EUV spectral range. Note also that the EUV and VUV data are taken for the same type of diode but not always for the same individual device.
Regarding radiation hardness, the change of the detector's spectral responsivity under irradiation can be divided in three main categories: growth of surface contamination layers; surface charging of insulating surface layers (such as SiO2) due to external photoelectron emission; and intrinsic damage (e.g., by creation of trapping states):
- 1. Growth of surface contamination layers: The detectors are used in vacuum because the radiation itself can only propagate in vacuum. The photon energy, on the other hand, is sufficiently high to ionize the residual gas atmosphere. The most abundant species in a common vacuum system (water and hydrocarbons) are cracked and the resulting radicals either stick to the surface and create carbon contamination or initialize oxidation in case of O
H. Nitridation of the SiO2–Si interface turned out to be advantageous for radiation hardness (Korde et al., 1993). For prolonged irradiation, however, these oxide-passivated diodes degrade extensively (Scholze et al., 2002, 2003). - 2. Surface charging of insulating surface layers (such as SiO2) due to external photoelectron emission: This issue was already addressed in the early investigations of VUV diodes (Korde and Geist, 1987), and changing from p-on-n to n-on-p polarity was suggested as a solution, because in this case the surface charging enhances the intrinsic field of the junction. Particularly, the formation of inhomogeneities was observed when irradiating photodiodes with a small beam (Fig. 7.7(a)) (Scholze et al., 1996). During an initial time period, the responsivity in the irradiated spot is stable but the region around the spot degrades. This was explained by a negative charging of the adjacent diode region due to photoelectrons accelerated back to the surface by the positive charge inside the irradiation spot and landing nearby on the oxide. This was proven by recovering the initial responsivity with a short irradiation of the initially shadowed area (Scholze et al., 1996).
- For the VUV spectral range, a pure-boron p-on-n diode was irradiated with a small beam. As discussed, for this polarity the positive charging of some parasitic oxide (probably grown in pinholes of the boron layer) decreases the responsivity. Fig. 7.8 shows the recovery of the responsivity over time because of a surface discharge process. This recovery, within a period of days, also indicates that the effect must be caused by rather small islands of oxide in a conductive matrix because the recovery is even slower for homogeneous oxide layers.
- In real devices, as shown in Fig. 7.2, the reverse contact forms an outer ring around the sensitive region and the inner contact ring for the face of the electrode. It is connected by deep-implanted doping to the reverse of the diode. The insulation between these two electrodes is usually achieved with SiO2. Irradiation of this region also charges the oxide, which, in this case, mainly results in an increased dark current (Shi et al., 2011).
- 3. Intrinsic damage (e.g., by creation of trapping states): This type of damage can be understood as a persistence of the initial impact of the primary process of charging (i.e., the creation of dangling bonds or other types of activated centers in the passivation layer of the diodes). This may result in permanent recombination centers or facilitate oxidation in the presence of OH radicals (i.e., residual water vapor in an unbaked vacuum system). The damage of the nitrided interface for a high irradiation dose could, indeed, be correlated to a radiation-induced oxidation of the surface as the K-absorption edge of O becomes more prominent in the spectral responsivity curve (Scholze et al., 1996). It is supposed that the combination of radiation-induced dangling bonds at the oxide–silicon interface in combination with OH radicals facilitates surface oxidation. Consequently, the oxide was substituted by an alternative passivation layer. Contamination was particularly investigated in the context of space-based VUV spectroscopy.
Figure 7.7 Effect of local surface charging by narrow photon beams on the responsivity of oxide-passivated n-on-p diodes (Scholze et al., 1996). The diode was irradiated in the center (position a) with a dose of the order of 10 mJ/mm2 at a wavelength of 9 nm. The radiation-induced charging of the oxide is positive in the spot (photoelectrons are ejected) and negative around the spot due to electrons back accelerated to the diode by the positive charge and landing nearby. The first measurement (●) was taken 3 months after this exposure, and the responsivity was normalized to the initial value. It was almost stable in the spot but decreased in the adjacent areas. This can be explained by the influence of the positive, respectively, negative charge on the responsivity of n-on-p diodes; here, a positive charge supports the charge collection because it attracts the electrons toward the n-layer at the surface. A second short exposure of 0.3 mJ/mm2 was placed at position b and a second scan was taken (○). In the irradiated spot, the responsivity was fully recovered by compensating the negative charge. In the adjacent areas, the responsivity was decreased by negative charging. In the position of the initial spot, the effect was less pronounced because the added negative charge was partly compensated by the remaining positive charge from the first exposure. This was proven by a third exposure at position c and a subsequent scan across the sensitive area (◇).
Figure 7.8 Vacuum ultraviolet exposure of a “pure-B” diode. Initially, the responsivity was homogeneous across the sensitive region (❍). The second scan was taken directly after a short exposure of only 0.04 mJ/mm2 with 121 nm radiation. The observed decrease in responsivity is almost 20% relative (◇). Here, the positive charge in the spot decreases the charge collection efficiency because the diode is p-on-n polarity. For this type of diode, the initial responsivity was almost recovered after 4 months storage of the device (△).
Whereas the changes due to categories 1 and 2 can, in principle, be reversed by suitable measures (cleaning, discharging), the processes due to category 3, lead to irreversible, permanent damage and, moreover, also influence the electrical properties of the device—typically, significantly raising the dark current (Shi et al., 2011) or increasing the surface recombination rate and, hence, leading to reduced responsivity (Gullikson et al., 1996). One common strategy to obtain higher radiation hardness regarding the latter is the use of a conductive (metallic) top layer. However, if this layer is simply added to the diode on top of the usual passivation layer, it substantially reduces the VUV responsivity due to its absorption. Therefore, the oxide passivation itself is replaced by the metallic top layer. At present, the front surfaces of devices with a metal-silicide passivation are significantly thicker than the thinnest oxide layer passivation; also, their passivation properties are inferior. Therefore, these devices have much lower responsivity in the VUV spectral range than diodes with oxide passivation. The latter, however, are not stable under VUV irradiation. Until now, only devices with a highly doped B-layer technique have combined radiation hardness with high responsivity (Aruev et al., 2009; Shi et al., 2010).
Summarizing, it can be stated that silicon diodes without an oxide passivation layer provide the best radiation hardness in the VUV and EUV spectral ranges. For the measurement of irradiation stability, it is possible to look directly for the signal change during irradiation. However, all the three categories above will always be observed in combination. To distinguish between them, it is necessary to perform measurements of spectral responsivity before and after irradiation, as well as after cleaning processes. For well-defined irradiation spots, this can also be achieved by scanning across the detector surface. In any event, these measurements need a high level of reproducibility to be comparable.
7.6. Future trends
The present status of VUV and EUV detectors based on the internal photoeffect can be summarized as follows:
- • The devices are usually silicon devices. Only for applications where high VUV sensitivity and simultaneously no responsivity in the UV and VIS are required (solar blindness), also wide-gap materials are in use.
- • All detectors with insulating top layers have severe issues with radiation hardness. At the moment, only highly boron-doped diodes achieve high VUV responsivity and radiation hardness simultaneously.
Future trends will therefore include work on the improvement of the metal-silicide passivation and the development of more reliable solar-blind wide-gap semiconductors, probably by applying the hybridization approach with silicon for those devices also.
For silicon devices, the appearance of ever-more functionalized detectors for different applications is directly on the horizon.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.