Photoreceivers today are in great demand for optical data storage applications. One of the most important subsystem components in a CD/DVD player, the optical pickup unit, requires a photodetector integrated circuit (PDIC). A PDIC combines a photodiode with the appropriate amplifier circuitry to read the information stored on optical media. The industry has turned to silicon-based semiconductor technologies for PDICs, integrating a silicon photodiode with a complementary metal-oxide semiconductor (CMOS) or BiCMOS amplifier.^{1 }

A key feature of silicon photodiode designs is the antireflective coating (ARC). The ARC maximizes the electrical responsivity of the device to a target optical wavelength by maximizing transmitted photons and minimizing reflected photons. As a result, engineers are actively developing ARCs of one or more layers with peak transmission at 780 nm (CD), 650 nm (DVD), and 405 nm (Blu-ray disc (BD, see p. 12)).

**Enhancing Performance** Photoreceiver sensitivity depends largely on the responsivity of its photodiode component. Defined as the ratio of generated photocurrent (amperes) to incident optical power (watts), the responsivity (R) figure of merit measured in amps per watt is a function of the structural and material characteristics of the photodiode and the transmission characteristics *T* of the ARC:

[1]

where *e/hc* is a physical constant, λ is the wavelength, η_{ext} is the external quantum efficiency, η_{int} is the internal quantum efficiency, and η_{c} is the collection efficiency.

When incident light impinges upon the device surface, a portion is reflected while the rest is transmitted and can be used for detection. In the absence of an ARC, the photodiode window is covered by thick oxide and nitride films for passivation, which can lead to worst-case reflection losses on the order of 50% (*T*~0.5); the reflection loss resulting from a simple air-to-silicon interface is on the order of 30% (*T*~0.7). Adding an ARC to minimize the reflection loss at the wavelength(s) of interest optimizes transmission.^{2} ARC design requires precise modeling of the reflection loss and is dependent on parameters such as materials, layer thicknesses and indices of refraction, and wavelength of operation.

**Calculating Reflectance** **Figure 1.** For analysis purposes, an ARC can be defined as an optical multilayer-thin-film system.

Based on multilayer-thin-film optics theory, engineers can create thin-film ARCs with virtually any desired reflectance and transmittance characteristics. To analyze these structures, one must consider the ARC as a series of thin films, each having a refractive index *n*_{j} and a thickness *d*_{j}. Light of a specified wavelength λ is incident normally from an absorption-free medium (typically air) and strikes the top layer (layer *M*_{1}) of a stack of films forming the total ARC. The last layer (*M*_{n}) is adjacent to the substrate (see figure 1).

To account for the interference effects of a system of thin films, one uses a ray approach with beams reflected backward and forward between the various interfaces. Light waves in each layer can be interpreted as two partial waves, one positive- and one negative-going. The laws of electromagnetic theory dictate that the tangential components of the electric and magnetic fields are continuous across a charge-free interface.^{3} Each layer making up the ARC is defined by a characteristic matrix *M*_{j }that relates the magnetic and electric fields at the layer's two interfaces and is a function of λ, *n*_{j}, and *d*_{j}. The matrix of the entire ARC (*M*_{ARC}) is the productin proper sequenceof these individual 2 x 2 characteristic layer matrices (*M*_{j }), as described by

[2]

where the layer's optical phase shift δ_{j} and optical admittance γ_{j }are given by

and

[3]

For *j* layers, there are ( *j*+1) boundaries or interfaces. The values of *E* and *M* represent the total electric and magnetic field amplitudes, respectively, at a particular frequency

[4]

where γ_{s} is the optical admittance of the substrate. The total reflectance (0 ¾ *R* ¾ 1) or reflection loss (*R*, in percent) of the multilayer stack or ARC in this case, is given in terms of *E *and *M *as

[5]

where *r* is the reflectivity and γ_{0} is the optical admittance of the incident medium. Consequently, only the characteristic ARC matrix (*M*_{ARC}) needs to be calculated and substituted in equation 5 to yield the ARC's reflectance.

The previously discussed model rests on several assumptions. First, the angle of incidence is assumed to be zero, as this is typical in CD/DVD system applications. Second, the model assumes negligible absorption by the ARC layers. Practically speaking, the cost-effective integration of a silicon photodiode in a CMOS or BiCMOS platform narrows down the available ARC materials to Si_{3}N_{4}, SiO_{2}, and polysilicon (poly-Si). All three are commonly deposited by plasma-enhanced chemical vapor deposition. One can also grow SiO_{2} thermally in an oxidizing ambient. Out of these three potential films, SiO_{2} and Si_{3}N_{4} are most widely used, while poly-Si absorbs some of the incident light.

**Table 1. **The table shows typical values for the refractive index (n) and region of transparency for silicon, SiO_{2}, and Si_{3}N_{4}.

This model assumes a uniform refractive index throughout each ARC layer. A material's refractive index can vary with wavelength and is usually specified within a particular range (see table). The choice of refractive indices will affect reflectance results. For our subsequent calculation-based results, we fixed the refractive indices at 2.0, 1.46, and 3.75 for Si_{3}N_{4}, SiO_{2}, and the silicon substrate, respectively.

**SLARCs and DLARCs** ARC designs compatible with metal-oxide semiconductor (MOS) process integration come in two common forms: a single layer ARC (SLARC) or a double layer ARC (DLARC). A SLARC works by producing two reflections that interfere destructively with each other. For the case of a single-thin-film ARC layer, equation 5 reduces to

[6]

Minima and maxima in SLARC reflectance occur for 90° and 0° optical phase shifts, respectively. The minimum reflectance *R*_{min} occurs for 90° optical phase shifts, corresponding to quarter-wavelength thickness (*d*_{1} = [R/4(*n*_{1})](*L*), where *L* is an odd integer),

[7]

SLARCs have generally used SiO_{2} because depositing Si_{3}N_{4} on silicon generates stress. One can achieve reflection-loss minima on the order of 7% if the SiO_{2} layer is engineered to be of λ/4 thickness. Nitride can form a very effective SLARC because it has a higher index of refraction than SiO_{2}, leading to a minimum reflection loss of ~0.1% at λ/4 thickness.

**Figure 2.** Plot shows reflection loss versus film thickness for a Si_{3}N_{4} SLARC at 650 and 780 nm.

CDs and current DVDs operate at 780 nm and 650 nm, respectively. Using the same ARC layer for photodiodes operating at both wavelengths is of great interest to PDIC manufacturers, but requires a SLARC yielding minimum reflection losses for dual-wavelength operation. Because 650-nm and 780-nm reflectance minima occur for film thicknesses within 10 to 20 nm of one another, an intermediate thickness can provide optimal performance (see figure 2).

The external quantum efficiency (QE) of a photodetector (PD) is much lower at shorter wavelengths than at longer ones due to the stronger absorption of light in silicon at the shorter wavelengths. The ARC layer optimization is thus more critical for a 405-nm PD than it would be for a 650-nm/780-nm PD, for example.^{4} While a SLARC may be sufficient for operation at wavelengths for which the QE is fairly high, a DLARC is critical for PDs operating at shorter wavelengths.

**Figure 3.** A 2-D contour plot shows reflectance of a DLARC for oxide thickness versus nitride thickness at 650- and 780-nm (left); the darker the region, the lower the reflectance. A DLARC combining 255 nm of oxide and 700 nm of nitride yields a near-zero reflection loss for both wavelengths (right).

For lower reflection loss at any wavelength, the best choice is a DLARC consisting of an SiO_{2} bottom layer and a Si_{3}N_{4} top layer, where the first reflection is canceled by interference with two weaker reflections.^{5} Because the refractive index of the top layer exceeds that of the bottom layer, reflection-loss minima will not be observed at standard quarter-wavelength thicknesses as in the SLARC example. For properly engineered layer thicknesses, DLARCs can achieve near-zero reflection losses. Plotting a 2-D reflectance contour map for top and bottom layer thicknesses reveals key combinations of layer thicknesses that yield minimum reflection losses at both 650 nm and 780 nm. A combination of a 255-nm-thick bottom oxide layer with a 700-nm-thick top nitride layer yields a near-zero reflection loss (see figure 3).^{6}

The optical storage market is driving the demand for photoreceivers built in a CMOS or BiCMOS technology, focusing attention on ARC designs that improve photodiode responsivity and photoreceiver sensitivity. For CD and current DVD technology, PDICs that use a Si_{3}N_{4} SLARC offer the simplest design with good performance. As wavelengths decrease, however, DLARCS are better suited because they can achieve near-zero reflection losses.

The standard for BD systems operating at 405 nm will require that they be able to read the 650-nm DVDs and 780-nm CDs for some time to come. In the near future, manufacturers will focus on ARC designs and integration schemes optimized for all three wavelengths of interest. New silicon-based materials, such as spin-on-glass-oxynitrides, silicon-oxynitrides, and polyimide thin films, which are compatible with MOS processing, could also lead to plausible ARC solutions. **oe**

*References*

*1. Ghazi, et al., Proc. of the 11th Intl. Meeting on Electro-Optics and Microelectronics in Israel, p. 228 (2000).*

*2. G. Bhaumik, et al., Journal of Physics of Semiconductor Devices 1, p. 607 (1998).*

*3. E. Hecht, Optics (second ed.), Addison-Wesley, Reading, MA (1987).*

*4. G. Thungstrom, et al., "Processing of silicon UV-photodetectors," Elsevier Science, p. 165 (2001).*

*5. Resnik, et al., IEEE, p. 1153 (1999).*

*6. K. Tsang, U.S. Patent #US2001/0011737 A1 (2001).*

**Phillip Espinasse, Steven Kosier**

*Phillip Espinasse is a product marketing engineer and Steven Kosier is vice president of engineering at PolarFab, Bloomington, MN. *