Physics and Technology - Vishay Semiconductors
Physics and Technology - Vishay Semiconductors
Physics and Technology - Vishay Semiconductors
Vishay Semiconductors
The best production method for all materials needed is liquid phase epitaxy (LPE). This method uses Ga-solutions containing As, possibly Al, and a doping substance. The solution is saturated at a high temperature, typically 900 C, and GaAs substrates are dipped into the liquid. The solubility of As and Al decreases with decreasing temperature. In this way epitaxial layers can be grown by slow cooling of the solution. Several layers differing in composition may be obtained using different solutions one after another, as needed e.g. for DHs. In liquid phase epitaxial reactors, production quantities of up to 50 wafers, depending on type of structure required, can be handled.
Au : Ge
Fig. 1
On an n-type substrate, two Si-doped layers are grown by liquid phase epitaxy from the same solution producing an emission wavelegth of 950 nm. Growth starts as n-type at high temperature and becomes p-type below about 820 C. A structured Al-contact on p-side and a large area Au:Ge contact on back side provide a very low series resistance. The angular distribution of emitted radiation is displayed in figure 2.
0 30
60
94 8197
20
40
60
80
90
Fig. 2 www.vishay.com 19
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
The package of the chip has to provide good collection efficiency of radiation emitted sideways, and has to diminish the refractive index step between the chip (n = 3.6) and the air (n = 1.0) with an epoxy of refractive index of 1.55. In this way, the output power of chip is increased by a factor of 3.5 for the assembled device. The chip described is the most cost-efficient one. Its forward voltage at IF = 1.5 A has the lowest possible value. Total series resistance is typically only 0.60 ; output power and linearity (defined as optical output power increase, divided by current increase between 0.1 A and 1.5 A) are high. Relevant data on chip and a typical assembled device are given in table 1. The technology used for a chip emitting at 880 nm eliminates the absorbing substrate and uses only a thick epitaxial layer. The chip is shown in figure 3.
Au : Ge/Au
30
60
94 8198
20
40
60
80
90
Fig. 4
n - GaAlAs : Si
Au : Zn
Due to its shorter wavelength, Ga1-XAlXAs chip described above offers specific advantages in combination with a Si detector. Integrated opto ICs, like amplifiers or Schmitt Triggers, have higher sensitivities at shorter wavelengths. Similarly, phototransistors are also more sensitive. Finally, the frequency bandwidth of pin diodes is higher at shorter wavelengths. This chip also has the advantage of having high linearity up to and beyond 1.5 A. The forward voltage, however, is higher than the voltage of a GaAs chip. Table 2 (see Symbols and Terminology) provides more data on the chip. A technology combining some of the advantages of the two technologies described above is summarized in figure 5.
p - GaAlAs : Ge p - GaAs : Si n - GaAs : Si
Fig. 3
Al
Originally, the GaAs substrate was adjacent to the n-side. Growth of Ga0.7Al0.3As started as n-type and became p-type - as in the first case - through the specific properties of the doping material Si. A characteristic feature of the Ga-Al-As phase system causes the Al-content of growing epitaxial layer to decrease. This causes the Al-concentration at the junction to drop to 8 % (Ga0.92Al0.08As), producing an emission wavelength of 880 nm. During further growth the Al-content approaches zero. The gradient of the Al-content and correlated gradient of bandgap energy produce an emission band of a relatively large half width. The transparency of the large bandgap material results in a high external efficiency on this type of chip. The chip is mounted n-side up, and the front side metallization is Au:Ge/Au, whereas the reverse side metallization is Au:Zn. The angular distribution of the emitted radiation is displayed in figure 4.
Au : Ge
Fig. 5
Starting an with n-type substrate, n- and p-type GaAs layers are grown in a similar way to the epitaxy of a standard GaAs:Si diode. After this, a highly transparent window layer of Ga1-XAlXAs, doped p-type is grown. The upper contact to the p-side is made of Al and the rear side contact is Au:Ge. The angular distribution of emitted radiation is shown in figure 6.
www.vishay.com 20
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
Vishay Semiconductors
30
Anode
p - GaAlAs p -GaAs
60
Cathode
94 8199
20
40
60
80
90
Fig. 6
Fig. 7
This chip type combines a relatively low forward voltage with a high electro-optical efficiency, offering an optimized combination of the advantageous characteristics of the two other chips. Refer again to table 2 (see Symbols and Terminology) for more details. As mentioned in the previous section, double heterostructures (DH) provide even higher efficiencies and faster switching times. A schematic representation of such a chip is shown in figure 7.
The active layer is depicted as the thin layer between the p- and n- type Ga1-XAlXAs confinement layers. The contacts are dependent on the polarity of the chip. If p is up, then the p-side contact is Al and the back side Au:Ge; if n is up, then this side has an Au:Ge contact and the back side Au:Zn.Two such chips that are also very suitable for IrDA applications are given in table 1.
Silicon Photodiodes (PN and PIN Diodes) The physics of silicon detector diodes Absorption of radiation is caused by the interaction of photons and charge carriers inside a material. The different energy levels allowed and the band structure determine the likelihood of interaction and, therefore, the absorption characteristics of the semiconductors. The long wavelength cutoff of the absorption is given by the bandgap energy. The slope of the absorption curve depends on the physics of interaction and is much weaker for silicon than for most other semiconducting materials. This results in a strong wavelength-dependent penetration depth which is shown in figure 8. (The penetration depth is defined as that depth where 1/e of the incident radiation is absorbed.)
Document Number: 80086 Rev. 1.0, 26-Aug-08
104 Si 103
100 101
102
102
0.6
1.0
94 8595
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
www.vishay.com 21
Depending on the wavelength, the penetration depth varies from tenths of a micron at 400 nm (blue) to more than 100 m at 1 m (IR). For detectors to be effective, an interaction length of at least twice the penetration depth should be realized (equivalent to 1/e2 = 86 % absorbed radiation). In the pn diode, generated carriers are collected by the electrical field of the pn junction. Effects in the vicinity of a pn junction are shown in figure 9 for various types and operating modes of the pn diode. Incident radiation generates mobile minority carriers - electrons on the p-side, holes on the n-side. In the short circuit mode shown in figure 9 (top), the carriers drift under the field of the built-in potential of the pn junction. Other carriers diffuse inside the field-free semiconductor along a concentration gradient, which results in an electrical current through the applied load, or without load, in an external voltage, open circuit voltage, VOC, at contact terminals. Bending of the energy bands near the surface is caused by surface states. An equilibrium is established between generation, recombination of carriers, and current flow through the load.
Surface States Without external bias
and circuits). The width, W, of the space charge is a function of doping the concentration NB and applied voltage V:
W =
2 S o ( V bi + V ) -------------------------------------------------------q NB
(1)
(for a one-sided abrupt junction), where Vbi is built-in voltage, s dielectric constant of Si, o vacuum dielectric constant and q is electronic charge. The diodes capacitance (which can be speed limiting) is also a function of the space charge width and applied voltage. It is given by
S o A C = --------------------------W
(2)
h Energy
94 8596
where A is the area of the diode. An externally applied bias will increase the space charge width (see figure 8) with the result that a larger number of carriers are generated inside this zone which can be flushed out very fast with high efficiency under the applied field. From equation (1), it is evident that the space charge width is a function of the doping concentration NB. Diodes with a so-called pin structure show according to equation (1) a wide space charge width where i stands for intrinsic, low doped. This zone is also sometimes nominated as n or p rather than low doped n, n- or p, p-zone indicating the very low doping. Per equation (2), the junction capacitance C, is low due to the large space charge region of PIN photodiodes. These photodiodes are mostly used in applications requiring high speed. Figure 10 shows a cross section of PIN photodiodes and PN diodes. The space charge width of the PIN photodiodes (bottom) with a doping level (n = NB) as low as NB = 5 x 1011 cm-3 is about 80 m wide for a 2.5 V bias in comparison with a pn diode with a doping (n) of NB = 5 x 1015 cm-3 with only 0.8 mm.
Contact Antireflection coating p+ Isolation layer
Space charge
1 to 10 cm
n+
94 8597
Fig. 9 - Generation-Recombination Effects in the Vicinity of a PN Junction Top: Short Circuit Mode, Bottom: Reverse Biased
Contact p+
94 8598
Recombination takes place inside the bulk material with technology- and process-dependent time constants which are very small near the contacts and surfaces of the device. For short wavelengths with very small penetration depths, carrier recombination is the efficiency limiting process. To achieve high efficiencies, as many carriers as possible should be separated by the electrical field inside the space charge region. This is a very fast process, much faster than typical recombination times (for data, see chapter Operating modes
www.vishay.com 22
Space charge
i n+
: to 10 000 cm
94 8599
Fig. 10 - Comparison of PN Diode (Top) and PIN Photodiode (Bottom) Document Number: 80086 Rev. 1.0, 26-Aug-08
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
Vishay Semiconductors
e = 0 I 1
V OC
I = I S ( exp V V T 1 )
with VT = kT/q k = 1.38 x 10-23 JK-1, Boltzmann constant q = 1.6 x 10-19 As, electronic charge.
(3)
94 8600
I SC
IS, the dark-reverse saturation current, is a material- and technology-dependent quantity. The value is influenced by the doping concentrations at pn junction, by carrier lifetime, and especially by temperature. It shows a strongly exponential temperature dependence and doubles every 8 C.
s ( ) = Ir e [ A W ]
(4)
I = I S ( exp V V T 1 ) s ( ) e
and in case V 0, zero or reverse bias we find,
Rsh = dV/dl
(5)
I = IS s ( ) e
(6)
80
Dependent on load resistance, RL, and applied bias, different operating modes can be distinguished. An unbiased diode operates in photovoltaic mode. Under short circuit conditions (load RL = 0 ), short circuit current, ISC flows into the load. When RL increases to infinity, the output voltage of the diode rises to the open circuit voltage, VOC, given by
V OC = V T ln ( s ( ) e I S + 1 )
(7)
The typical dark currents of Si photodiodes are dependent on size and technology and range from less than picoamps up to tens of nanoamps at room temperature conditions. As noise generators, dark current Ir0 and the resistance Rsh (defined and measured at a voltage of 10 mV forward or reverse, or peak-to-peak) are limiting quantities when detecting very small signals. The photodiode exposed to optical radiation generates a photocurrent Ir exactly proportional to incident radiant power e.
Because of this logarithmic behavior, the open circuit voltage is sometimes used for optical light meters in photographic applications. The open circuit voltage shows a strong temperature dependence with a negative temperature coefficient. The reason for this is the exponential temperature coefficient of the dark reverse saturation current IS. For precise light measurement, a temperature control of the photodiode is employed. Precise linear optical power measurements require small voltages at the load, typically smaller than about 5 % of the corresponding open circuit voltage. For less precise measurements, an output voltage of half the open circuit voltage can be allowed. The most important disadvantage of operating in photovoltaic mode is the relatively large response time. For faster response, it is necessary to implement an additional voltage source reverse-biasing the photodiode. This mode of operation is termed photoconductive mode. In this mode, the lowest detectable power is limited by the shot noise of the dark current, IS, while in photovoltaic mode, the thermal (Johnson) noise of shunt resistance, Rsh, is the limiting quantity.
www.vishay.com 23
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
1.0
0.8
0.6
I re = N q s = I re e
=
(8) (9)
N q (h N) = q (h )
94 8602
Wavelength (nm)
( m ) s = ---------------- [ A W ] 1.24
At fixed efficiency, a linear relationship between wavelength and spectral responsivity is valid. Figure 8 shows that the semiconductors absorb radiation similar to a cut-off filter. At wavelengths smaller than the cut-off wavelength, the incident radiation is absorbed. At larger wavelengths the radiation passes through the material without interaction. The cut-off wavelength corresponds to the bandgap of the material. As long as the energy of the photon is larger than the bandgap, carriers can be generated by absorption of photons, provided that the material is thick enough to propagate photon-carrier interaction. Bearing in mind that the energy of photons decreases with increasing wavelength, we can see, that the curve of the spectral responsivity vs. wavelength in ideal case (100 % efficiency) will have a triangular shape (see figure 13). For silicon photodetectors, the cut-off wavelength is near 1100 nm. In most applications, it is not necessary to detect radiation with wavelengths larger than 1000 nm. Therefore, designers use a typical chip thickness of 200 m to 300 m, which results in reduced sensitivity at wavelengths larger than 950 nm. With a typical chip thickness of 250 m, an efficiency of about 35 % at 1060 nm is achieved. At shorter wavelengths (blue-near UV, 500 nm to 300 nm) sensitivity is limited by recombination effects near the surface of the semiconductor. A reduction in efficiency starts near 500 nm and increases as the wavelength decreases. Standard detectors designed for visible and near IR radiation may have poor UV/blue sensitivity and poor UV stability. Well designed sensors for wavelengths of 300 nm to 400 nm can operate with fairly high efficiencies. At shorter wavelengths (< 300 nm), efficiency decreases strongly.
Fig. 13 - Spectral Responsivity as a Function of Wavelength of a Si Photodetector Diode, Ideal and Typical Values
Temperature dependence of spectral responsivity The efficiency of carrier generation by absorption and the loss of carriers by recombination are the factors which influence spectral responsivity. The absorption coefficient increases with temperature. The radiation of the long wavelength is therefore more efficiently absorbed inside the bulk and results in increased response. For shorter wavelengths (< 600 nm), reduced efficiency is observed with increasing temperature because of increased recombination rates near the surface. These effects are strongly dependent on technological parameters and therefore cannot be generalized to the behavior at longer wavelengths. Uniformity of spectral responsivity Inside the technologically defined active area of photodiodes, spectral responsivity shows a variation of sensitivity on the order of < 1 %. Outside the defined active area, and especially at lateral edges of the chips, local spectral response is sensitive to applied reverse voltage. Additionally, this effect depends on wavelength. Therefore, the relation between power (W) related spectral responsivity, s (A/W), and power density (W/cm2) related spectral responsivity, s [A/(W/cm2)] is not a constant. Rather, this relation is a function of wavelength and reverse bias Stability of spectral responsivity Si detectors for wavelengths between 500 nm and 800 nm appear to be stable over very long periods of time. In the literature concerned here, remarks can be found on instabilities of detectors in blue, UV, and near IR under certain conditions. Thermal cycling reversed the degradation effects. Surface effects and contamination are possible causes but are technologically well controlled. Angular dependence of responsivity The angular response of Si photodiodes is given by the optical laws of reflection. The angular response of a detector is shown in figure 14.
www.vishay.com 24
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
Vishay Semiconductors
1.0 A/W
30
0.8
0.6 Acm2/W
60
0.4
0.2 0 - 90
94 8603
- 60
- 30
0.2
0.4
0.6
0.8
90
Angle
Relative Responsitivity
Semiconductor surfaces are covered with quarter wavelength anti-reflection coatings. Encapsulation is performed with uncoated glass or sapphire windows. The bare silicon response can be altered by optical imaging devices such as lenses. In this way, nearly every arbitrary angular response can be achieved. Dynamic Properties of Si Photodiodes Si photodiodes are available in many different variations. The design of diodes can be tailored to meet special needs. Si photodiodes may be designed for maximum efficiency at given wavelengths, for very low leakage currents, or for high speed. The design of a photodiode is nearly always a compromise between various aspects of a specification. Inside the absorbing material of the diode, photons can be absorbed in different regions. For example at the top of a p+n--diode there is a highly doped layer of p+ - Si. Radiation of shorter wavelengths will be effectively absorbed, but for larger wavelengths only a small amount is absorbed. In the vicinity of the pn junction, there is the space charge region, where most of the photons should generate carriers. An electric field accelerates the generated carrier in this part of the detector to a high drift velocity. Carriers which are not absorbed in these regions penetrate into field-free region where the motion of the generated carriers fluctuates by a slow diffusion process. The dynamic response of the detector is composed of different processes which transport carriers to contacts. The dynamic response of photodiodes is influenced by three fundamental effects: Drift of carriers in an electric field Diffusion of carriers Capacitance x load resistance Carrier drift in the space charge region occurs rapidly with very small time constants. Typically, transit times in an electric field of 0.6 V/m are on the order of 16 ps/m and 50 ps/m for electrons and holes, respectively. At (maximum) saturation velocity, the transit time is on the order
of 10 ps/m for electrons in p-material. With a 10 m drift region, travelling times of 100 ps can be expected. Response time is a function of the distribution of the generated carriers and is therefore dependent on wavelength. The diffusion of the carriers is a very slow process. Time constants are on the order of some ms. The typical pulse response of the detectors is dominated by these two processes. Obviously, carriers should be absorbed in large space charge regions with high internal electrical fields. This requires material with an adequate low doping level. Furthermore, a reverse bias of rather large voltage is useful. Radiation of shorter wavelength is absorbed in smaller penetration depths. At wavelengths shorter than 600 nm, decreasing wavelength leads to an absorption in the diffused top layer. The movement of carriers in this region is also diffusion limited. Because of the small carrier lifetimes, the time constants are not as large as in homogeneous substrate material. Finally, capacitive loading of output in combination with load resistance limits frequency response.
n n n+
94 8605
p EA
Backside contact
The switching times of phototransistors are dependent on current amplification and load resistance and are between 30 ms and 1 ms. The resulting cut-off frequencies are a few hundred kHz.
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]
www.vishay.com 25
t r, f =
( 1 2f t ) + b ( RC B V )
2 2
(10)
ft: R: CB: V:
Phototransistors are most frequently applied in transmissive and reflective optical sensors.
www.vishay.com 26
For technical questions concerning emitters, contact: [email protected] For technical questions concerning detectors, contact: [email protected]