hat contains G-H-As PT Ee dRice&: Photodetectors “The detector is like the journalist who must determine what, where, when, which, and how? What is the identity of the particle? Exactly where is it when it is observed? When does the particle get to the detector? Which way is it going? How fast is it moving?” Sheldon L. Glashow' {A selection of commercial InGaAs based photodetectors, including fiber-pigtalled pho- ‘odiodes. (Courtesy of Fermianics, California.) 5.1 PRINCIPLE OF THE pV JUNCTION PHOTODIODE Photodetectors convert a light signal to an electrical signal such as a voltage or current. In many photodetectors such as photoconductors and photodiodes this conversion is, typically achieved by the creation of free electron hole pairs (EHPs) by the absorption of photons, that is, the creation of electrons in the conduction band (CB) and holes in the valence band (VB). In some devices such as pyroelectric detectors the energy con- version involves the generation of heat which increases the temperature of the device which changes its polarization and hence its relative permittivity. We will consider pr "Sheldon L-Glashow Imeractions (Warner Books, New York, 1988),p. 1015.2 Ramo's Theorem and External Photocurrent 219, space charge density py. in Figure 5.1 (b) across W subject to a voltage difference of V, + V,~The field only exists in the depletion region and is not uniform. It varies across the depletion region as shown in Figure 5.1 (c) where it is maximum at the junction and penetrates into the n-side. The regions outside the depletion layer are the neutral regions in which there are majority carriers. It is sometimes convenient to treat these neutral re- sions simply as resistive extensions of electrodes to the depletion layer. When a photon with an energy greater than the bandgap E, is incident, it becomes. absorbed to photogenerate a free EHP, that is an electron in the CB and a hole in the ‘VB. Usually the energy of the photon is such that photogeneration takes place in the de~ pletion layer. The field E in the depletion layer then separates the EHP and drifts them in opposite directions until they reach the neutral regions as depicted in Figure 5.1 (a), Drifting carriers generate a current, called photocurrent /,,,in the external circuit that provides the electrical signal. The photocurrent lasts for the duration it takes ‘or the electron and hole to cross the depletion layer (W’) and reach the neutral regions. When the drifting hole reaches the neutral p*-region it recombines with a electron entering the p'-side from the negative electrode, that is from the battery. Similarly, when the drift- ing electron reaches the neutral n-side, an electron leaves the n-side into the electrode (battery). The photocurrent J, depends on the number of EHPs photogenerated and the drift velocities of the carriers while they are transiting the depletion layer. Since the field is not uniform and the absorption of photons occurs over a distance that depends on the wavelength, the time dependence of the photocurrent signal cannot be deter- mined in a simple fashion. It should be mentioned that the photocurrent in the external circuit is due to the flow of electrons only even though there are both electrons and holes drifting within the device. Suppose that there are N number of EHPs photogenerated. If we were to inte- arate the photocurrent /,, to calculate how much charge has flowed we would find a charge Q that is due to the total number of photogenerated electrons (eN) and not due to both electrons and holes (2e/V), 52 RAMOS THEOREM AND EXTERNAL PHOTOCURRENT Consider a semiconductor material with a negligible dark conductivity that is electrod- ed and biased as shown in Figure 5.2 (a). The electrodes do not inject carriers but allow excess carriers in the sample to leave and become collected by the battery (they are termed noninjecting electrodes). The field £ in the sample is uniform and itis V /L. We will ater see that this situation is almost identical to the intrinsic region of a reverse bi- ased pin photodiode. Suppose that a single photon is absorbed at a position x = [from the left electrode and instantly creates an electron hole pair. The electron and the hole drift in opposite directions with respective drift velocities v, = 1, and V, = yE,where uand jx, are the electron and hole drift mobilities respectively. The transit time of a car- Tier is the time it takes for a carrier to drift from its generation point to the collecting electrode. The electron and hole transit times f, and 1, respectively are marked on the t s.r diagram in Figure 5.2 (b) where, Eo! md et 0 t Elecion and hoe220 Chapter Photodetector: FIGURE 5.2 (2) An EHP is photogenerated at x in opposite directions with drift velocities vy and v,.(b) Consider first only the drifting electron. Suppose that the external phot due to the motion of this electron is i,(1). The electron is acted on by the force electric field. When it moves a distance dx, work must be done by the ex In time di, the electron drifts a distance dx and does an amou provided by the battery in time de as Vi,(t) dt. Thus, Work done It is apparent that this current continues to flow as (has a velocity ,) in the sample. It lasts for a duration ¢, at the end of which the elec tron reaches the battery. Thus, although the clectron has been phot taneously, the external photocurrent is nor instantaneous and has generated instan time spread. Figur We can apply similar arguments to the drifting hole as well which will generat nole photocurrent j fhe external circuit given, as in Figure 5.2 (c), by5.3 Absorption Coefficient and Photodiode Materials 221 The total external current will be the sum of i,(¢) and i,(¢) as shown in Figure 52) If we integrate the external current i,4(¢) to evaluate the collected charge Quseact we wank ad, Quateaes = [ “ode + if “ig tae ‘This result can be verified by evaluating the area under the i,y(¢) curve in Figure 52 (d). Thus, the collected charge is not 2e but just one electron as shown by the area in Figure 5.2 (d). Equations (2) to (4) constitute Ramo’s theorem. In general, if charge q is being drifted with a velocity v,(t) by a field between two biased electrodes separat- ed by L. then this motion of q generates an external current given by wat), erat: ‘The total external current is the sum of all currents of the type in Eq. (5) from all rifting charges between the electrodes. e @ 1 < bent «) ame i) 5.3 ABSORPTION COEFFICIENT AND PHOTODIODE MATERIALS ‘The photon absorption process for photogeneration, that is the creation of EHPs, re- quires the photon energy to be at least equal to the bandgap energy E, of the semicon- ductor material to excite an electron from the valence band (VB) to the conduction band (CB). The upper cut-off wavelength (or the threshold wavelength) A, for photo- tocurrent Seah generative absorption is therefore determined by the bandgap energy E, of the semi- conductor so that h(c/A,) = E, oF, Cue 124 wavelengths Alum) = Bay) yn Kendo For example, for Si, E, = 1.12 €V and A, is 1.11 um whereas for Ge E, = 0.66eV and the corresponding A, ~ 1.87 um. tis clear that Si photodiodes cannot be used in ‘optical communications at 1.3 and 1.55 um whereas Ge photodiodes are commercially available for use at these wavelengths. Table 5.1 lists some typical bandgap energies and the corresponding cut-off wavelengths of various photodiode semiconductor materials. Incident photons with wavelengths shorter than A, become absorbed as they trav- élin the semiconductor and the light intensity, which is proportional to the number of photons, decays exponentially with distance into the semiconductor. The light inter ata distance x from the semiconductor surface is given by K(x) = L,exp(—ax) @ ere where fs the intensity ofthe ineident radiation and a is the absorption coefficient that depends on the photon energy or wavelength A. Absorption coefficient a is a material property. Most of the photon absorption (63%) occurs over a distance 1 /a and I fa is222 Chapter5—_Photodetectors Tay Gays 08 InAs 035 Insb 018 FIGURE 5.3 The absorption efficient (a) vx. wavelength (A) fr ‘arious semiconductors (data se 10 lectvely collected and combined Systems, Second Ed, (Prem bao? Hall, New York, 193), p44; Melchior "Demodulation a: Photodetection Techniques” ie EI Arecohi and A.O, Schult Dubois, Eds Laser Handbook ow!) Vol. 1 (North-Holland, Amster: dam, 1972), pp. 725-835, 08 called the penetration depth 5. Figure 5.3 shows the a vs. A characteristics of various is apparent that the behavior of a with the wavelength A de pends on the semiconductor material. In direct bandgap semiconductors such as III-V semiconductors (e.g. GaAs, InAs InP, GaSb) and in many of their alloys (e.g. InGaAs, GaAsSb) the photon absorption, process is a direct process that requires no assistance from lattice vibrations. The pho- ton is absorbed and the electron is excited directly from the valence band to the con duction band without a change in its k-vector (or its crystal momentum fik) inasmuch as the photon momentum is very small. The chan; valence to the conduction band fiken — fikyn = photon momentum ~ 0. This pro corresponds to a vertical transition on the E-k diagram, that is electron energy (E) vs. in the electron momentum from the5.3. Absorption Coefficient and Photodiode Materials 223, E £ i (©) Giahs Direct bandgap) (6) Si lnivect bandgap) FAGURE 5.4 (a) Photon absorption in a direct bandgap semiconductor. (b) Photon absorption in an ind rest handgap semiconductor (VB, valence band; CB, conduction band) electron momentum (fik) in the erystal in Figure 5.4 (a). The absorption coefficient of these semiconductors rises sharply with decreasing wavelength from A, as apparent for GaAs and InP in Figure 5.3. In indirect bandgap semiconductors such as Si and Ge, the photon absorption for photon energies near E, requires the absorption and emission of lattice vibrations, that isphonons,’ during the absorption process as shown in Figure 5.4 (b). If Kis the wavev: torofalattice wave (lattice vibrations travel inthe crystal) then 7K represents the mo- mentum associated with such a lattice vibration, that is #K is a phonon momentum. When an electron in the valence band is excited to the conduction band there isa change inits momentum in the erystal and this change in the momentum cannot be supplied by the momentum of the incident photon which is very small. Thus, the momentum differ- ence must be balanced by a phonon momentum: Bikey — ltkyy = phonon momentum = AK. The absorption process is said to be indirect as it depends on lattice vibrations which in turn depend on the temperature. Since the interaction of a photon with a va- lence electron needs a third body, a lattice vibration, the probability of photon absorp- tion is not as high as ina direct transition. Furthermore, the cut-off wavelength is not as, sharp as for direct bandgap semiconductors. During the absorption process, a phonon may be absorbed or emitted. If # is the frequency of the lattice vibrations then the phonon energy is hd. The photon energy is hy where v is the photon frequency. Con- servation of energy requires that hw = a) ‘Thus, the onset of absorption does not exactly coincide with E,, but typically itis very close to E, inasmuch as h# is small (<0.1 eV). The absorption coefficient initially "As much as an electromagnetic radiation is quantized in terms of photons lattice vibrations in the nal are quantized in terms of phonons.224 ChapterS —_Photodetectors rises slowly with decreasing wavelength from about A, as apparent in Figure 5.3 for Ge and Si The choice of material for a photodiode must be such that the photon energies are greater than E,. Further, at the wavelength of radiation, the absorption occurs over a depth covering the depletion layer so that the photogenerated EHPs can be separat ed by the field and collected at the electrodes. If the absorption coefficient is too larg then absorption will occur very near the surface of the p* layer which is outside the de pletion layer. First, the absence of a field means that the pho rated electron can only make it to the depletion layer to cross to the n-side by diffusion. Secondly, photo generation near the surface invariably leads to rapid recombination due to surface d: ts that act as recombination centers. On the other hand, if the absorption coefficient {s too small, only a small portion of the photons will be absorbed in the depletion layer and only a limited number of EHPs can be photogenerated, 5.4 QUANTUM EFFICIENCY AND RESPONSIVITY Not all the incident photons are absorbed to create free electron-hole pairs (EHPs) th can be collected and give rise to a photocurrent. The efficiency of the conversion process of received photons to free EHPs is measured by the quantum efficiency (QE) 7 of the detector defined as Number of free EHP generated and collected n i) Number of incident photons The measured photocurrent J, in the external citcuit is due to the flow of electrons per second to the terminals of the photodiode. Number of electrons collected per se ond is J,,/e. If P, is the incident optical power then the number of photons arriving per second is P,/hv. Then the QE 7 can also be defined by Ile f 1 Pie a Not all of the absorbed photons may photogenerate free EHPs that can be col: lected. Some EHPs may disappear by recombination without contributing to the photocurrent or become immediately trapped. Further if the semiconductor length is comparable with the penetration depth (1/a) then not all the photons will be ab sorbed. The device QE is therefore always loss than unity. It depends on the ab: sorption coefficient a of the semiconductor at the wavelength of interest and on the structure of the device. QE can be increased by reducing the reflections at the semi: conductor surface, increasing absorption within the depletion layer and preventing the recombination or trapping of carriers before they are collected. The QE defined in Eq. (1) is for the whole device. More specifically, it is known as the external quan im efficiency. Internal quantum efficiency is the number of free EHPs photogen erated per absorbed photon and is typically quite high for many devices, The QE ‘Free implies caries that can be coll jction band and hol5.5 Thepin Photodiode 225 definition in Eq. (1) incorporates the internal quantum efficiency as it applies to the whole device ‘The responsivity R of a photodiode characterizes its performance in terms of the photocurrent generated (J,«) per incident optical power (P,) at a given wavelength, Photocurrent (A) = - 3) > Theident Optical Power (W) ® From the definition of QE, itis clear that, e_ ek Uae “ In Eq, (4) n depends on the wavelength. The responsivity therefore clearly d pends on the wavelength. R is also called the spectral responsivity or radiant sensitiv ty. The R vs. A characteristics represents the spectral response of the photodiode and is generally provided by the manufacturer. Ideally with a quantum efficiency of 100% (1 = 1),R should increase with A up to A, as indicated in Figure 5.5. In practice, QE limits the responsivity to lie below the ideal photodiode line with upper and lower wave- length limits as shown for a typical Si photodiode in the same figure. The OE of a well designed Si photodiode in the wavelength range 700-900 nm can be close to 90-95! 5.5 THE pin PHOTODIODE ‘The simple pn junction photodiode (Figure 5.1) has two major drawbacks Its junction or depletion layer capacitance is not sufficiently small to allow photodetection at high modulation frequencies. This is an RC time constant limitation. Secondly, its depletion layer is at most a few microns. This means that at long wavelengths where the penetra~ tion depth is greater than the depletion layer width, the majority of photons are ab: sorbed outside the depletion layer where there is no field to separate the EHPs and drift them. The OE is correspondingly low at these long wavelengths. These problems are substantially reduced in the pin (p-intrinsic-n-type) photodiode.” FIGURE 5.5 _Responsviy (R) wavelength (A) for an ideal photo diode with QE = 100% (4 = 1) and for atypical commercial Si Photodiode, Resposivty (AW) 0" 200" 400” 60” Sb "10" 1300 Wavelength (om) “The pin photodiode was invented by J. Nishizawa and his esearch group in Japan in 1950. Responsisity and OE226 Chi FIGURE 5.6 ap rs Photodetectors ture ofan idealized pin photodiode. Elect ) The et spece charge den p across the photodiode. () The built in field across the diode. (d) The pin P ions a a @ © | e © Thi schematically illustrated in the idealized structure in Figure 5.6 (a). The intrinsic layer hhas a much smaller doping than both p* and n* regions and it is much wider than these in refers to a semiconductor device that has the structure p*-intrinsic-n" as regions, typically 5-50 um depending on the particular application, In the idealized pin photodiode we can take, for simplicity, the i-Si region to be truly intrinsic, When the structure is first formed, holes diffuse from the p"-side and electrons from n'-side into the i-Silayer where they recombine and disappear. This leaves behind a thin layer of ex posed negatively charged acceptor ions in the p”-side and a thin layer of exposed posi 5.6 (b). The two charges are separated by the /-Si layer of thickness W. There is a uniform built-in field €, in -Si layer from the exposed positive ions to exposed negative ions as illustrated in Figure 5.6 (c). In contrast, the built-in field in the depletion layer of a pn junction is not uniform. With tively charged donor ions in the -side as shown in Fi fie Th ph ph to} to th ti5.5 Thepin Photodiode 227 10 applied bias, equilibrium is maintained by the built-in field E, which prevents further diffusion of majority carriers into the /-Si layer. The separation of two very thin layers of negative and positive charges by a fixed distance, width W of the i-Si, is the same as that in a tion or depletion layer capacitance of the pin diode is given by ee, _ hts Cy = Oy where A is the cross sectional area and e,6, isthe permittivity of the semiconductor (Gi). Further, since the width W of the i-Si layer is fixed by the structure, the junction capacitance does not depend on the applied voltage in contrast to that of the pn junc- tion. Cj, is typically of the order of a picofarad in fast pin photodiodes so that with a 50 0 resistor, the RC,,» time constant is about 50 ps. When a reverse bias voltage V, is applied across the pin device, it drops almost en tirely across the width of i-Si layer. The depletion layer widths of the thin sheets of ceptor and donor charges in the p* and n” sides are negligible compared with W. The erse bias V, increases the built-in voltage to V, + V, as shown in Figure 5.6 (d). The field £ in the /Si layer is still uniform and increases to Vv, > Vv) @ “np cscs Svipin os Gh ps tcp os cee nga The photogenerated EHPs inthe Si layer are then separated by the field F and drift: bd towards Che n° endp* sien reopetsively as Wtnted in’ gure 5.6 (4). Wale the pctopcocrated carriers ae Glie Garoigh The £54 lay thoy he ae 06 ot ntcraal photocurrent which is detected as a voltage across a small resistor Rin Figure 5.6 (2). Tc response lime of the pin photodiode determined by the transit mes of the pho- iogeneratedcariersacros the width W ofthe /Silaer. Increasing W allows more Pho- tons tobe absorbed which increases the OE but it slows down the speed of response as cact transit times become longer. For a charge carvier that is photogenerated atthe edge on the FSi layer the transit time or it ime age across the #Si layer i tan = = ® where visits drift velocity. To reduce the drift time, that is increase the speed of response, we have to increase v, and therefore increase the applied field E. At high fields v, does not follow the expected .,£ behavior, where 4, is the drift mobility, but instead tends tosaturate at Vy, Which is of the order of 10°m at fields greater than 10° min the «ase of Si. Figure 5.7 shows the variation ofthe drift velocity of electrons and holes with the field in Si. The v; = jE behavior is only observed at low fields. At high fields both the drift velocities saturate, For an i-Si layer of width 10 um, with carriers driftn Ssturation velocities the drift time is about 0.1 ns which is longer than typical RCy,» time constants. The speed of pin photodiodes are invariably limited by the transit time of photogenerated carriers across the i-Si ayer. sapactance opin Biased pin