2.2.1 The Ideal Crystal

A large single crystal of pure silicon (Figure 2.3) forms the starting point for the monocrystalline silicon solar cell—the most efficient type in common use. As we shall see, the simple and elegant structure of such crystals makes it comparatively easy to explain the basic semiconductor physics and operation of PV cells. We are talking here of silicon refined to very high purity, similar to that used by the electronics industry to make semiconductor devices (diodes, transistors, and integrated circuits including computer chips). Its purity is typically 99.99999%. This contrasts with the far less pure metallurgical‐grade silicon, produced by reducing quartzite in electric arc furnaces, that is used to make special steels and alloys.

The Czochralski (CZ) method of growing silicon crystals is quite easy to visualize. Chunks of pure silicon with no particular crystallographic structure are melted at 1414°C in a graphite crucible. A small seed of silicon is then brought into contact with the surface of the melt to start crystallization. Molten silicon solidifies at the interface between seed and melt as the seed is slowly withdrawn. A large ingot begins to grow both vertically and laterally with the atoms tending to arrange themselves in a perfect crystal lattice.

Unfortunately, this classic method of producing crystals has a number of disadvantages. Crystal growth is slow and energy intensive, leading to high production costs. Impurities may be introduced due to interaction between the melt and the crucible. And in the case of PV, the aim is of course to produce thin solar cell wafers rather than large ingots, so wire saws are used to cut the ingot into thin slices, a time‐consuming process that involves discarding valuable material. For these reasons the PV industry has spent a lot of R&D effort investigating alternatives, including pulling crystals in thin sheet or ribbon form, and some of these are now used in volume production. Whatever method is employed, the desired result is pure crystalline silicon with a simple and consistent atomic structure.

Semiconductors, such as Si, are made up of individual atoms bonded together in a structure where each atom is surrounded by eight electrons. The electrons surrounding each atom of Si are part of a covalent bond, consisting of two atoms “sharing” a single electron. Each Si atom forms four covalent bonds with the four surrounding atoms. This is illustrated in Figure 2.4(a). Each line connecting the atoms represents an electron being shared between the two. Since each atom has four valence electrons that are not tightly bound to its nucleus, a perfect lattice structure is formed when each atom forms bonds with its four nearest neighbors (which are actually at the vertices of a three‐dimensional tetrahedron, but shown here in two dimensions for simplicity). The structure has profound implications for the fundamental physics of silicon solar cells.

Silicon in its pure state is referred to as an intrinsic semiconductor. It is neither an insulator like glass nor a conductor like copper, but something in between. At low temperatures its valence electrons are tightly constrained by bonds, as in part (a) of the figure, and it acts as an insulator. But bonds can be broken if sufficiently jolted by an external source of energy such as heat or light, creating electrons that are free to migrate through the lattice. If we shine light on the crystal, the tiny packets, or quanta, of light energy can produce broken bonds if sufficiently energetic. The silicon becomes a conductor, and the more bonds are broken, the greater its conductivity.

Figure 2.4(b) shows an electron ε₁ that has broken free to wander through the lattice. It leaves behind a broken bond, indicated by a dotted line. The free electron carries a negative charge and, since the crystal remains electrically neutral, the broken bond must be left with a positive charge. In effect it is a positively charged particle, known as a hole. We see that breaking a bond has given rise to a pair of equal and opposite charged “particles,” an electron and a hole. At first sight the hole might appear to be an “immovable object” fixed in the crystal lattice. But now consider the electron ε₂ shown in the figure, which has broken free from somewhere else in the lattice. It is quite likely to jump into the vacant spot left by the first electron, restoring the original broken bond, but leaving a new broken bond behind. In this way a broken bond, or hole, can also move through the crystal, but as a positive charge. It is analogous to people sitting in a theater row and one by one moving one seat to the side. The moving people resemble the electrons and the emptied seats resemble the holes that appear as moving in the opposite direction.

We see that the electrical properties of intrinsic silicon depend on the number of mobile electron–hole pairs in the crystal lattice. At low temperatures, in the dark, it is effectively an insulator. At higher temperatures, or under sunlight, it becomes a conductor. If we attach two contacts and apply an external voltage using a battery, current will flow—due to free electrons moving one way and holes on the other. We have now reached an important stage in understanding how a silicon wafer can be turned into a practical solar cell.

Yet there is a vital missing link: remove the external voltage and the electrons and holes wander randomly in the crystal lattice with no preferred directions. There is no tendency for them to produce current flow in an external circuit. A pure silicon wafer, even under strong sunlight, cannot generate electricity and become a solar cell. What is needed is a mechanism to propel electrons and holes in opposite directions in the crystal lattice, forcing current through an external circuit and producing useful power. This mechanism is provided by one of the great inventions of the 20th century, the semiconductor p–n junction.

2.2.2 The p–n Junction

A conventional monocrystalline solar cell has a silvery top surface surmounted by a fine grid of metallic fingers forming one of its electrical contacts. What is less obvious is that the cell actually consists of two different layers of silicon that have been deliberately doped with very small quantities of impurity atoms, often phosphorus and boron, to form a p–n junction. The addition of such dopants is absolutely crucial to the cell’s operation and provides the mechanism that forces electrons and holes generated by sunlight to do useful work in an external circuit.

The p–n junction may be regarded as the basic building block of the semiconductor revolution that began back in the 1950s. It is perhaps a little surprising that an invention normally associated with mainstream electronics should also form the basis of PV technology; but a silicon solar cell is essentially a form of p–n junction specially tailored to the task of converting sunlight into electricity.

We have already noted that heating or shining light on pure silicon can alter its electrical properties, progressively converting it from an insulator into a conductor. Another extremely important way of modifying its properties is to add small amounts of dopants. For example, if phosphorus is added to molten silicon, the solidified crystal contains some phosphorus atoms in place of silicon. While the latter has four valence electrons able to form bonds with neighboring atoms, phosphorus has five. The extra one is only weakly bound to its parent atom and can easily be enticed away, as shown in Figure 2.5(a). In other words, silicon doped with phosphorus provides plenty of free electrons, known as the majority carriers. Generally, there are also a few holes present due to thermal generation of electron–hole pairs, as in intrinsic silicon, and these are called minority carriers. The material is a fairly good conductor and is referred to as negative‐type or n‐type.

A complementary situation arises if silicon is doped with boron, which has only three valence electrons loosely bound to its nucleus, illustrated in part (b) of the figure. Each boron atom can only form full bonds with three neighboring silicon atoms, so boron introduces broken bonds into the crystal. In this case holes are the majority carriers and electrons the minority carriers. Once again, the material becomes a conductor; it is referred to as positive‐type or p‐type.

We see that n‐type material has many surplus electrons and p‐type material has many surplus holes. The next step is to consider what happens when the two materials are joined together to form a p–n junction, illustrated in Figure 2.6(a).

Near the interface, free electrons in the n‐type material start diffusing into the p‐side, leaving behind a layer that is positively charged due to the presence of fixed phosphorus atoms. Holes in the p‐type material diffuse into the n‐side, leaving behind a layer that is negatively charged by the fixed boron atoms. This diffusion of the two types of majority carriers, in opposite directions across the interface, has the extremely important effect of setting up a strong electric field, creating a potential barrier to further flow. Equilibrium is established when the tendency of electrons and holes to continue diffusing down their respective concentration gradients is offset by their difficulty in surmounting the potential barrier. In this condition there are hardly any mobile charge carriers left close to the junction and a so‐called depletion region is formed.

The depletion region makes the p–n junction into a diode, a device that conducts current easily in one direction only. Figure 2.6(b) shows an external voltage V applied to the diode, making the p‐type material positive with respect to the n‐type, referred to as forward bias. In effect the external voltage counteracts the “built‐in” potential barrier, reducing its height and encouraging large numbers of majority carriers to cross the junction—electrons from the n‐side and holes from the p‐side. This results in substantial forward current flow (note that conventional positive current is actually composed of negatively charged electrons flowing the other way; we may think of them as going right around the circuit through the battery and back into the n‐type layer). Conversely if the external voltage is inverted to produce a reverse bias, the potential barrier increases and the only current flow is a very small dark saturation current (I₀). This is because a bias that increases the potential barrier for majority carriers decreases it for minority carriers—and at normal temperatures there are some of these present on both sides of the junction due to thermal generation of electron–hole pairs.

The practical result of these movements of electrons and holes is summarized by the diode characteristic in Figure 2.7. Diode current I increases with positive bias, growing rapidly above about 0.6 V; but with negative bias the reverse current “saturates” at a very small value I₀. Clearly this device only allows current flow easily in one direction. Mathematically the curve is expressed as

(2.1)

where q is the charge on an electron, k is Boltzmann’s constant (1.3807 × 10⁻²³ J/K), and T is the absolute temperature (K).

Equation (2.1) is called the diode or Shockley equation.

You are perhaps beginning to wonder what all this has to do with solar cells, because we have not so far discussed the effects of shining light on the diode and it is not obvious what these will be. However, rest assured that understanding the aforementioned discussion of electrons and holes, majority and minority carriers, and potential barriers is essential for unraveling the mysteries of PV!

2.2.3 Monocrystalline Silicon

2.2.3.1 Photons in Action

We are now close to understanding how a monocrystalline silicon wafer, doped to create a semiconductor diode, can work as a power‐generating solar cell. The basic scheme in Figure 2.8 shows a small portion of such a cell. At the top, several metallic contact fingers form part of the cell’s negative terminal. Next comes a thin layer of n‐type material interfacing with a thicker layer of p‐type material to produce the crucial p–n junction. And finally there is a back contact that acts as the positive terminal. For clarity the cell’s thickness is exaggerated in the figure; it is actually a very slim wafer, normally less than 0.3 mm from top to bottom.

A stream of photons containing minute packets or quanta of energy shines on the cell. Their numbers are staggering: under strong sunlight a 6‐inch (15‐cm) cell receives more than 10¹⁹ photons every second. Various possible fates await them, some productive and others fruitless, and we show a few important examples in the figure.

Unfortunately, there is some loss of photons by optical reflection back from the conducting fingers, top surface, and rear surface (nos. 1, 2, and 3 in the figure). The rest enter the cell body, but only those with a certain minimum energy, known as the bandgap, have any chance of creating an electron–hole pair and contributing to the cell’s electrical output. The most productive ones, for reasons explained in the following text, create electron–hole pairs in the p‐type layer or in the n‐type layer very close to the junction (4 and 5). Less productive, on average, are the ones that travel further into the p‐type material (6). Successful cell design involves producing as many electron–hole pairs as possible, preferably close to the junction. But even high‐quality cells are subject to theoretical limits dictated by the spectral distribution of sunlight, nature of light absorption in silicon, and quantum theory. We shall discuss these topics a little later.

First comes the big question: what happens to the electron–hole pairs generated within the cell by sunlight, and how do they produce current flow in an external circuit?

As we have seen, majority carriers (electrons in n‐type material, holes in p‐type) are the main players in a conventional semiconductor diode. By initial diffusion across the p–n junction, they set up a depletion layer and create a potential barrier. Forward‐biasing the diode reduces the height of the barrier, making it easier for them to cross the junction and produce substantial current. In reverse bias the barrier increases and current flow is severely inhibited. Diode action is principally due to the behavior of majority carriers under the influence of an applied external voltage.

With solar cells, however, it is light‐generated minority carriers that take center stage in creating electric current. The basic reason may be simply stated: a potential barrier that inhibits transfer of majority carriers across a p–n junction positively encourages the transfer of minority carriers. Whereas majority carriers experience “a hill to climb,” minority carriers see “a hill to roll down.” With luck they are swept down this hill, collected at the cell terminals, and produce an output current proportional to the intensity of the incident light.

Let us consider the three photons in Figure 2.8 that successfully create electron–hole pairs in the crystal lattice. Number 4 produces a pair in the p‐type region, close to the junction. Its free electron, a minority carrier in p‐type material, is easily swept across the junction and collected. So is the hole produced in the n‐type region by number 5, which is swept across the junction in the opposite direction. Both these minority carriers should contribute to the light‐generated current.

Photon 6 also creates an electron–hole pair, but well away from the junction and its associated electric field. The free electron does not immediately experience “a hill to roll down,” but instead starts wandering randomly through the silicon lattice. In the figure it is shown eventually reaching the junction and being swept away to success. But the journey is a dangerous one: it may instead encounter a hole and be annihilated. Although such recombination is not illustrated in the figure, unfortunately it occurs not only in the main body of the cell (bulk recombination) but also even more importantly at the edges and metal contacts due to defects and impurities in the crystal.

The longer a minority carrier wanders around, the greater the distance traveled through the crystal and the more likely it is to be lost by recombination. Two measures are used to describe the risk. The carrier lifetime is the average amount of time between electron–hole generation and recombination (the bigger, the better), which for silicon is typically 1 µs. The diffusion length is the average distance a carrier moves from the point of generation until it recombines, for silicon typically 0.2 mm that is comparable with the thickness of the monocrystalline wafer. This again emphasizes the value of electron–hole pairs generated close to the junction.

We have now covered some fundamental aspects of solar cell operation, including the key role played by light‐generated minority carriers. The next task is to consider the voltage–current characteristics of the cell as measured at its output terminals.

2.2.3.2 Generating Power

We have seen solar photons at work, creating minority carriers that speed toward the solar cell’s output terminals under the magical influence of the p–n junction. But how is all this internal activity reflected in the cell’s power generation, and what voltages and currents are produced at its terminals? Figure 2.9(a) helps answer the question with an equivalent circuit summarizing the cell’s behavior as a circuit component. It consists of a diode representing the action of the p–n junction together with a current generator representing the light‐generated current I_L.

In dark conditions I_L is zero and the cell is quiescent. If an external voltage source is connected, the cell behaves just like a semiconductor diode with the characteristic shown in part (b) of the figure (this has the same form as Figure 2.7). We choose to define the current I as flowing into the circuit and, in the dark, it must be the same as the diode current I_D. Note also that since a diode is a passive device that dissipates power, the cell’s dark characteristic lies entirely in the first and third quadrants (I and V either both positive or both negative). But if sufficient sunlight falls on the cell to turn it into an active device delivering power to the outside world, the current I must reverse and the characteristic will shift into the fourth quadrant (I negative, V positive) shown shaded in the figure.

In sunlight the generator produces a current I_L proportional to the level of insolation. It is effectively superimposed on the normal diode characteristic, and we may write

(2.2)

Substituting for the diode current using Equation (2.1) gives

(2.3)

This equation confirms that the diode I–V characteristic is shifted down into the fourth quadrant by an amount equal to the light‐generated current I_L. This is shown in Figure 2.10(a).

Most people are unfamiliar with curves in the fourth quadrant, so for convenience the I–V characteristics of a solar cell are normally “flipped over” to the first quadrant. This is equivalent to plotting V against −I. Part (b) of the figure illustrates a family of such curves for a typical crystalline silicon cell rated by the manufacturer at 2 W_p. Each curve represents a different strength of sunlight and hence a different value of I_L. You will recall that PV cells and modules are normally rated in peak watts (W_p), indicating the maximum power they can deliver under standard conditions (insolation 1000 W/m², cell temperature 25°C, AM1.5 solar spectrum). Therefore, we should first consider how the rated power of 2 W_p relates to the 1000 W/m² I–V curve.

In general, the cell’s power output equals the product of its voltage and current. No power is produced on open circuit (maximum voltage, zero current) or short circuit (maximum current, zero voltage). The full rated power is obtained by operating the cell slightly below maximum voltage and current at its maximum power point (MPP), shown as P₁ against the 1000 W/m² curve, and corresponding to about 4 A at 0.5 V or 2 W. We can only obtain the promised output power by operating the cell at its MPP. Three other curves are shown for lower insolation values of 750, 500, and 250 W/m²; each has its own MPP (P₂, P₃, P₄), indicating the maximum power available from the cell at that particular strength of sunlight.

Note that the maximum voltage produced by a silicon solar cell is about 0.6 V, considerably less than the 1.5 V of a dry battery cell. This means that it is essentially a low‐voltage, high‐current device, and many cells must be connected in series to provide the higher voltages required for most applications. For example, the PV module previously illustrated in Figure 2.1 has 72 individual cells connected in series, giving a DC voltage of about 35 V at the MPP. Higher voltages may be obtained by connecting a number of modules in series.

The I–V characteristics suggest another important aspect of the solar cell—it is helpful to think of it as a current source rather than a voltage source like a battery. A battery has a more or less fixed voltage and provides variable amounts of current; but at a given insolation level, the solar cell provides a more or less fixed current over a wide range of voltage.

The maximum voltage of the cell, its open‐circuit voltage V_oc, is given by the intercept on the voltage axis and lies in the range 0.5–0.6 V. It does not depend greatly on the insolation. The close relationship between the diode characteristic of the p–n junction and the I–V characteristics in sunlight, illustrated in Figure 2.10(a), means that the open‐circuit voltage is similar to the forward voltage of about 0.6 V at which a silicon diode starts to conduct heavily.

The maximum current from the cell, its short‐circuit current I_sc, is given by the intercept on the current axis and is proportional to the strength of the sunlight. Other things being equal, it is also proportional to the cell’s surface area. It represents the full flow of minority carriers generated by the sunlight and successfully “collected” after crossing the p–n junction.

The aforementioned parameters are further illustrated in Figure 2.11. The blue curve shows a typical I–V characteristic at 1000 W/m² insolation, labeled with the short‐circuit current, open‐circuit voltage, and MPP. The red curve shows how power output varies with voltage; the maximum value is P_mp = I_mp × V_mp. Since the current holds up well over most of the voltage range, it follows that the cell’s output power is roughly proportional to voltage up to the MPP. This emphasizes once again the importance of operating the cell close to the MPP if its power output potential is to be realized.

A widely used measure of performance that reflects the overall quality of the cell is its fill factor (FF) given by

(2.4)

An “ideal” cell in which the current held right up to the short‐circuit value, then reduced suddenly to zero at the MPP, would have an FF of unity. Needless to say, practical cells do not achieve this; the I–V characteristics in the figure have an FF of about 70%. Equation (2.4) shows that graphically it is equal to the ratio between the areas of the small and large shaded rectangles in the figure.

So far we have not considered the effects of temperature on cell performance, but actually they are quite important, especially in the case of crystalline silicon. Many people imagine that solar cells are more efficient if operated at elevated temperatures, perhaps thinking of the type of solar–thermal panel used for water heating. But solar PV cells like to be kept cool—they do very well in strong winter sunshine in the Swiss Alps! In hot climates cell temperatures can reach 70°C or higher and system designers often go to considerable lengths to ensure adequate ventilation of PV modules to assist cooling.

The main effect of temperature on a cell’s I–V characteristics is a reduction in open‐circuit voltage, illustrated in Figure 2.12. We have repeated the 1000 W/m² curve for the 2 W_p cell already shown in Figure 2.10(b) for the standard temperature of 25°C and added two further curves for 0 and 50°C. The open‐circuit voltage changes by about 0.1 V between these extremes, corresponding to 0.33% per °C. Note that the temperature coefficient is negative; in other words the voltage decreases as the temperature rises. There is a much smaller effect on the short‐circuit current. Generally, the cell loses power at elevated temperatures, a more serious effect with crystalline silicon than most other types of solar cell.

You have probably noticed one major omission from this discussion—an explanation of efficiency. At the start of this chapter, we noted that commercial multicrystalline silicon modules have typical efficiencies in the range 13–16%, but we have not so far explained the reasons for this apparently rather disappointing performance. Returning for a moment to Figure 2.10(b), it is not clear from our discussion why this cell, which probably receives up to about 14 W_p of incident solar energy, only manages to convert 2 W_p into electrical output. Where does the rest go, and why cannot the efficiency be dramatically improved by better design? This raises some fundamental issues that we tackle in the next section.

2.2.3.3 Sunlight, Silicon, and Quantum Mechanics

It may seem a little surprising to find “quantum mechanics” mentioned in an introductory book on PV—and possibly unnerving in view of a quotation by Richard Feynman (1918–1988), latterly a professor at the California Institute of Technology, who received a Nobel Prize in Physics in 1965 for his work on quantum mechanics and famously declared: “I think I can safely say that nobody understands quantum mechanics.”

So it is clear we must tread lightly, leaving the great body of 20th‐century quantum theory undisturbed. Yet not entirely, for it contains precious nuggets relating to the nature of sunlight and imposes fundamental limits on the efficiency of solar cells.^1,2

Back in Section 1.4 we noted that certain eminent physicists, from Isaac Newton in the 17th century to Albert Einstein in the 20th, viewed light as a stream of minute particles carrying discrete packets of energy. And in Section 2.2.3.1 we stated—without explanation—that a light quantum or photon needs to have a certain minimum energy, known as the bandgap, if it is to have any chance of creating an electron–hole pair in a silicon crystal lattice. It is now time to bring these ideas together with the help of a little quantum theory.

The human eye is sensitive to visible light—all the colors of the rainbow from violet to red. The corresponding range of wavelengths is about 0.4–0.8 µm. The complete solar spectrum, previously shown in Figure 1.6, also contains significant energy at ultraviolet (UV) and especially infrared (IR) wavelengths. A key concept of quantum theory is that the energy content of a photon is related to wavelength by a surprisingly simple equation:

(2.5)

where E is the photon energy, h is Planck’s constant (6.62607004 × 10⁻³⁴ m² kg/s), c is the velocity of light (~3 × 10⁸ m/s), and λ is the wavelength. This means that the packet of energy or quantum is about twice as large for a violet photon as for a red photon. And as Einstein proposed in 1905, quanta can only be generated or absorbed as complete units.

When E is expressed in electron volts (eV) and λ in micrometers (µm),

(2.6)

A second key point is that solar cells based on semiconductors are essentially quantum devices. An individual solar photon can only generate an electron–hole pair if its quantum of energy exceeds the bandgap of the semiconductor material, also known as its forbidden energy gap. This is illustrated in Figure 2.13 and the energy bandgaps of various semiconductors are shown in Figure 2.14.

You may recall that the creation of an electron–hole pair involves jolting a valence electron to produce a broken bond in the crystal lattice. The electron moves from the valence band to the conduction band, leaving behind an equal, but oppositely charged hole. However, the energy levels of an electron in the two bands are separated by a discrete energy gap. Moving from one band to another requires a “quantum leap”—it is all or nothing, and intermediate levels are forbidden. Long‐wavelength IR and red photons do not generally have the necessary amount of energy. Conversely most photons toward the violet end of the spectrum have more than enough and the excess must be dissipated as heat. These fundamental considerations, taken in conjunction with the sun’s spectral distribution, reduce the theoretical maximum efficiency of a silicon solar cell at an insolation of 1000 W/m² to about 30%. The figure does not take account of various other loss mechanisms and practical design considerations, some of which were illustrated in Figure 2.8. So it is not hard to appreciate why cells made in research laboratories do well to reach 30% and why most commercial, mass‐produced cells achieve less than 20%.

The Shockley–Queisser limit of 30% shown in Figure 2.14 corresponds to (i) single junction, (ii) unconcentrated light, (iii) T = 25°C, and (iv) thermal relaxation so that energy above the bandgap is dissipated as thermal energy instead of contributing to energizing more electrons. To go above the Shockley–Queisser, we need to use concentrated light, more than one junction so a stack of cells one at the top of the other, and/or materials and configurations that can give better quantum efficiency.

We can now appreciate why the size of the bandgap is a very important influence on solar cell efficiency. It represents the minimum energy needed to free an electron from its valence band to the conduction band so that it can move within the crystal and conduct electricity. The energy bandgap of semiconductors tends to slightly decrease as the temperature is increased because interatomic spacing increases with the amplitude of atomic vibrations increasing due to the increased thermal energy. An increased interatomic spacing decreases the potential seen by the electrons in the material, which in turn reduces the energy bandgap. The levels shown in Figure 2.14 correspond to 0°C (273.15°K).

It becomes obvious now why there are several types of solar cells based on different semiconductor materials. Using a semiconductor with a wide bandgap yields a device with a high voltage, but photons are “thrown away.” Using a semiconductor with a narrow bandgap yields a device with a high current (as more photons are absorbed) but low voltage (Figure 2.15). Since P = V × I, there is not an optimal bandgap that corresponds to the maximum product of current and voltage, although most efficient harvesting of the sun’s energy requires bandgaps in the range 1.0–1.6 electron volts (eV). Silicon’s bandgap of 1.1 eV is fairly good in this respect. Certain other semiconductor materials (e.g., CdTe, GaAs) have bandgaps closer to the middle of the range, and we will discuss them later.

Unfortunately, not all photons with the necessary energy are readily absorbed. Most solar cell materials, the direct‐bandgap semiconductors, act as good light absorbers within layers just a few micrometers thick. But crystalline silicon, an indirect‐bandgap material, is not so effective. It absorbs high‐energy blue photons quite easily, close to the cell’s top surface, but low‐energy red photons generally travel much further before absorption and may exit the cell altogether. The basic problem is that successful generation of conduction electrons in silicon requires additional quantum lattice vibrations that complicate the process, so that layers less than about 0.1 mm thick are not good light absorbers. Special light‐trapping techniques may be used to increase the path length of light inside the cell and give a better chance of electron–hole generation. These are described in the next section.

To summarize, it would be helpful if every photon entering a solar cell produced an electron–hole pair and contributed to power generation, in other words if the quantum efficiency, thus the number of electrons divided by the number of photons, was 100%. But quantum theory tells us this is impossible. Photons are all‐or‐nothing packets of energy that can only be used in their entirety. Some are too feeble in their energy content, while others are unnecessarily strong, placing fundamental limits on solar cell efficiency. Disappointing though this may seem, we should always remember that sunlight is “free” energy to be used or not as we wish. Photons are not wasted if untapped—at least not in the sense of an old‐fashioned power station burning fossil fuel that effectively discards around 60% of its precious fuel as waste heat.

2.2.3.4 Refining the Design

Solar cell designers are constantly striving to improve conversion efficiencies and have used their ingenuity over many years to refine crystalline silicon cells beyond the basic scheme already illustrated in Figure 2.8. Some of the constraints on efficiency are caused by fundamentals of light and quantum theory, others by the properties of semiconductor materials or the problems of practical design.

One important point should be made at the outset. Researchers use various sophisticated techniques to achieve “record” efficiencies and can select their best cells for independent testing and accreditation. But PV companies engaged in large‐scale production have an additional set of priorities: simple, reliable, and rapid manufacturing processes and high yield coupled with minimal use of expensive materials, all aimed at lower costs. Manufacturers are certainly interested in the commercial advantages of high cell efficiency and over the years have incorporated many design advances coming out of research laboratories, but cost must always be a big consideration and there are often significant time lags.

Figure 2.16 summarizes the main factors determining the efficiency of a typical, commercial, crystalline silicon solar cell operated at or near its MPP. On the left the incident solar power is denoted by 100%. Successive losses, shaded in blue, reduce the available power to around 15–20% at the cell’s output terminals—its rated efficiency value. We will now discuss each loss category in turn.

Quantum Theory

We emphasized the fundamental limitations imposed by quantum theory³ in the previous section. They represent the biggest loss of efficiency in a solar cell based on a single p–n junction. One way of reducing the problem is to stack together two or more junctions with different bandgaps, creating a tandem cell. A well‐known example, which has been exploited commercially for many years, is based upon amorphous rather than crystalline silicon, and we shall mention this again in Section 2.3.

Optical Losses

Optical losses affect the incoming sunlight, preventing absorption by the semiconductor material and production of electron–hole pairs. The small section of solar cell shown in Figure 2.17 illustrates three main categories of optical loss: blocking of the light by the top contact (i); reflection from the top surface (ii); and reflection from the back contact without subsequent absorption (iii).

Shadowing by the top contact can obviously be minimized by making the total contact area as small as possible. This area comprises not only the metallic contact fingers shown in the figure (and previously in Figure 2.8) but also wider strips known as bus bars that join many fingers together and conduct current away from the cell. Clearly a well‐spaced grid of very fine fingers and narrow bus bars helps reduce optical loss, but the disadvantage is increased electrical resistance. As always, practical design involves compromise.

The photo in Figure 2.18 shows the top surface of a monocrystalline silicon cell, surrounded by its neighbors in a PV module. This example has very simple grid geometry, consisting of 49 fine vertical fingers and two horizontal bus bars, giving a shadowing loss of about 11%. The fingers have constant width; a more efficient design would taper them to account for the increasing current each carries as it nears a bus bar. The bus bars are slightly tapered toward the low‐current end; it would be better to taper them along their length as they pick up current from more and more fingers. Ideally the cross sections of fingers and bus bars should be roughly proportional, at each point, to the current carried. To illustrate this, a small section of a more efficient finger–bus bar design is shown in part (b) of the figure.

The metallization pattern of fingers and bus bars, as well as having its own inherent resistance to current flow, introduces contact resistance at the semiconductor interface. This may be reduced by heavy doping of the top layer of semiconductor material, at the risk of forming a significant dead region at the surface that reduces the collection efficiency of blue photons.

Conventional top contacts are made from very thin metallic strips formed using a screen‐printing process. A metallic paste is squeezed through a mask, or screen, depositing the desired contact pattern that is then fired. The shading loss, typically between 8 and 12%, represents a significant drain on cell efficiency. A major design improvement, pioneered in the 1990s at the University of New South Wales,⁴ uses laser‐formed grooves to define a metallization pattern with narrower but deeper fingers just below the cell’s surface. Such buried contact solar cells offer valuable gains in efficiency compared with normal screen‐printed designs.

The second category of optical loss illustrated in Figure 2.17 is reflection from the cell’s top surface. Two main design refinements are commonly employed. The first is to apply a transparent dielectric antireflection coating (ARC) to the top surface, illustrated in Figure 2.19. If the coating is made a quarter‐wavelength thick, the light wave reflected from the ARC/silicon interface is 180° out of phase with that reflected from the top surface, and when the two combine, the resulting interference effects produce cancelation. This condition is met when

(2.7)

where d is the thickness, n is the refractive index of the coating material, and λ is the wavelength (interestingly, we are temporarily considering light as a wave rather than a stream of particles, a good example of the dual nature of light first mentioned in Section 1.4). Clearly, exact cancelation can only occur at one value of λ, normally chosen to coincide with the peak photon flux about 0.65 µm. The antireflection performance falls off to either side of this value. For optimum performance the refractive index of the ARC material should be intermediate between that of the materials on either side, usually silicon or either air or glass.

The second design refinement involves texturizing the top surface so that light is reflected in a fairly random fashion and has a better chance of entering the cell. Almost any roughening is helpful, but the crystalline structure of silicon offers a special opportunity because careful surface etching can be used to create a pattern of minute raised pyramids, illustrated in Figure 2.20. Light reflected from the inclined pyramidal faces is quite likely to strike adjacent pyramids and enter the cell.

The third type of optical loss is reflection of light from the back of the cell, without subsequent absorption. This may be reduced by an uneven back surface that reflects the light in random directions, trapping some of it in the cell by total internal reflection. The technique is referred to as light trapping⁴ and is very important in crystalline silicon cells because silicon is a relatively poor light absorber, especially of longer‐wavelength (red) light. It is illustrated in Figure 2.21.

It is difficult to put precise figures on the efficiency losses caused by these various optical effects. However, a cell that includes carefully designed metallization, ARC, texturization, and light trapping can give major improvements compared with the basic structure first illustrated in Figure 2.8.

Recombination Losses

The undesirable process known as recombination has already been discussed in Section 2.2.3.1. It occurs when light‐generated electrons and holes, instead of being swept across the p–n junction and collected, meet up and are annihilated. The wastage of charge carriers adversely affects both the voltage and current output from the cell, reducing its efficiency.

Some recombination takes place as electrons and holes wander around in the body of the cell (bulk recombination), but most occurs at impurities or defects in the crystal structure near the cell’s surfaces, edges, and metal contacts, as illustrated in Figure 2.22. The basic reason is that such sites allow extra energy levels within the otherwise forbidden energy gap (see Figure 2.13). Electrons are now able to recombine with holes by giving up energy in stages, relaxing to intermediate energy levels before finally falling back to the valence band. In effect they are provided with stepping stones to facilitate the quantum leaps necessary for recombination.

What can be done to reduce recombination? Three important techniques may be briefly mentioned here. The first involves processing the cell to create a back surface field (BSF). Although the details are subtle,³ the tendency of long wavelength photons to recombine at the back of the cell may be reduced by including a heavily doped aluminum region, which also acts as the back contact. Next, it is possible to reduce recombination at the external surfaces by chemical treatment with a thin layer of passivating oxide. And finally, regions adjacent to the top contacts may be heavily doped to create “minority carrier mirrors” that dissuade holes in the n‐type top layer from approaching the contacts and recombining with precious free electrons.

Resistance Losses

The final efficiency loss shown in Figure 2.16 is due to electrical resistance. We previously noted that a solar cell is best thought of as a current generator. As with other current generators, it is desirable to minimize resistance in series with the output terminals and maximize any shunt resistance that appears in parallel with the current source. A low shunt resistance, which could happen due to manufacturing defects, would cause power losses in solar cells by providing an alternate current path for the light‐generated current.5 Figure 2.23 shows two equivalent circuits similar to that previously used for a solar cell (Figure 2.9) but modified to include a series resistance R₁ and a shunt resistance R₂. Ideally, R₁ would be zero and R₂ infinite, but, needless to say, we cannot expect these values in practice.

The physical interpretation of R₁ is straightforward. It represents the resistance to current flow offered by the bus bars, fingers, contacts, and the cell’s bulk semiconductor material. A well‐designed cell keeps R₁ as small as possible. R₂ is more obscure, relating to the nonideal nature of the p–n junction and impurities near the cell’s edges that tend to provide a short‐circuit path around the junction. In practical designs both resistors cause losses, but it is simpler to appreciate their effects if we treat them separately.

The I–V, also termed J–V, characteristic shown in (a) is for R₁ = 0 and R₂ = infinity, which is the ideal case. Part (b) in Figure 2.23 shows characteristics for a cell with finite values of R₁ and R₂. Let us first consider the open‐circuit condition, I = 0. In this case there is no current through R₁ and no voltage drop across it, so the open‐circuit voltage V_oc must be the same as for the reference cell. We assumed that series resistance due to a cell’s bus bars, fingers, contacts, and semiconductor material has no effect on the open‐circuit voltage. However, full circuit analysis shows that it causes a small reduction in short‐circuit current and a loss of FF, as indicated. R₁ (R_s) causes a slight rotation of the I–V curve around the (V_oc, I = 0) point.

To consider the effects of shunt resistance, it is helpful to consider the short‐circuit condition, V = 0. In this case there is no voltage across R₂ and no current through it, so the short‐circuit current I_sc must be the same as for the reference cell. Thus, a finite shunt resistance due to imperfections in and around the cell’s p–n junction has no effect on the short‐circuit current. However, it has a minor effect on the open‐circuit voltage and a considerable one on the FF. The shunt resistance R₂ (r_sh) causes a slight rotation of the I–V curve around the (I_sc, V = 0) point. To conclude, a practical cell with both series and shunt resistance losses is expected to suffer small reductions in both V_oc and I_sc; but the most serious effect is generally degradation of the FF (defined in Equation (2.4)).

This interdependency can be quantified by replacing the simplified model shown in Figure 2.9 and Equation (2.3) by a more standard model that includes both the series resistance R₁ and the shunt resistance R₂/(shown in Figure 2.23). Then for deriving the characteristic I–V curves, the current becomes I = I_l − I_D − I₂, where I₂ is estimated from Equation (2.7):

(2.8)

Then the characteristic I–V equation is given as

(2.9)

We have now covered the main categories of efficiency loss in crystalline silicon solar cells. The techniques for counteracting them have been conceived and enhanced over many years in R&D laboratories around the world, leading to continuous improvements in cell and module efficiencies. Of course, the degree to which they are employed in a commercial product depends upon the manufacturer’s expertise and judgment; the number and complexity of processing steps have a big impact on cost and there is inevitably a trade‐off between cost and performance.

2.2.4 Multicrystalline Silicon

In most respects multicrystalline silicon, also referred to as polycrystalline silicon or more simply as poly‐Si, produces solar cells that are very similar to their monocrystalline cousins. The theoretical background is shared, even though the initial stage of manufacture is different. As first mentioned in Section 2.1, multicrystalline cells also start life as pure molten silicon, but the material is cast in substantial blocks, cut into smaller bricks, and finally made into thin wafers. The casting process produces a multi‐grain crystal structure that is less ideal than monocrystalline material and gives cell and module efficiencies typically 1% (absolute) lower, but this disadvantage is offset by lower wafer costs. And since the cells are cut square or rectangular, rather than “pseudo‐square” as with monocrystalline cells, they can be packed closely in modules. They have a scaly, shimmering appearance. The façade exhibited in Figure 2.2 shows that the modules tend to have a distinctly blue appearance due to their ARC, a property often appreciated by architects.

As the molten silicon cools, crystallization occurs simultaneously at many points, producing crystal grains with random sizes, shapes, and orientations. After cutting into thin wafers, the material has the surface appearance in Figure 2.24(a). Within each grain the crystal structure is highly regular, but the many grain boundaries represent imperfections and provide unwelcome sites for electron–hole recombination. The problem is reduced if grains are at least a few millimeters across and extend from front to back of the wafer. As part (b) of the figure shows, a multicrystalline module tends to present a uniform, shimmering appearance without the gaps between cells associated with the “pseudo‐square” shape of monocrystalline cells.

On the whole there is little to choose between the performance of monocrystalline and multicrystalline PV modules. From a user’s point of view, efficiency and cost differences may not be decisive and the choice often comes down to appearance, availability, and the manufacturer’s reputation and guarantee.

2.3.1 Amorphous and Thin‐Film Silicon

Amorphous silicon (a‐Si) was the first thin‐film technology used in PV. Small a‐Si cells in consumer products such as watches and calculators have introduced solar cells to millions of people since the 1980s. The tiny amounts of power required by such products make the comparatively low efficiency of their cells unimportant, and in any case they are rarely used outdoors under strong sunlight! Ease of manufacture and low cost are their strong points. What is not so generally realized is that a‐Si technology has been developed in recent years and scaled up for higher‐power applications. Although it only accounts for a few percent of world production, it is no longer confined to consumer products. A good example is building façades; a‐Si modules can serve as attractive cladding and may well be competitive with other types of PV module. PV cladding is not necessarily more expensive than traditional high‐quality materials and may be chosen for its aesthetic appeal or as an environmental statement. If the façade also generates electricity, so much the better. Efficiency is not the only criterion.

In any case the question of efficiency needs further discussion. We noted at the start of this chapter that a‐Si module efficiencies typically fall in the range 6–9%, about half that of crystalline silicon. But efficiencies quoted by PV manufacturers invariably relate to standard insolation (1000 W/m², 25°C) and tell only part of the story. While crystalline silicon modules are impressive under strong sunlight, their performance in weak or diffuse light is often inferior to thin‐film products and is more adversely affected by high temperatures. In recent years there have been many reports of thin‐film modules, both a‐Si and CdTe, outperforming crystalline silicon in terms of annual energy yield, especially in climates with significant cloud cover and plenty of diffused light.

Amorphous silicon is also a far better light absorber than crystalline silicon, so extremely thin layers of semiconductor may be used—of the order 1 µm. Like other thin‐film technologies, it offers more advantages:

• Relatively simple fabrication at low temperatures using inexpensive substrates and continuous “production line” methods
• Integrated, monolithic design obviating the need to cut and mount individual wafers
• Potential for manufacturing flexible, lightweight products

The word amorphous, derived from ancient Greek, means “without form or shape.” a‐Si, which may be deposited as a thin film on a variety of substrates, does not exhibit a regular lattice structure. The distances and angles between the silicon atoms are randomly distributed, giving rise to incomplete bonds and a high concentration of defects. The result is a high density of allowed energy states within the nominal energy gap, in stark contrast to crystalline silicon. In effect, the extra energy states act as stepping stones, allowing conduction electrons to relax back into the valence band and recombine. There is also a problem of low charge‐carrier mobility within the semiconductor material (referred to as poor carrier transport). Fortunately, early research into a‐Si solar cells suggested two effective ways of countering these difficulties.

First, it was discovered that introducing hydrogen into amorphous silicon could passivate incomplete bonds, also known as dangling bonds (DBs), greatly reducing the number of excess energy states within the bandgap. The modified material is referred to as a‐Si(H) to denote its hydrogen content and is illustrated in Figure 2.26. This shows the irregular arrangement of silicon atoms, a DB, and a DB that has been passivated by a hydrogen atom (H). Using this approach it is possible to make effective n‐type and p‐type materials by doping with phosphorus or boron, resulting in a direct bandgap semiconductor with an energy gap of about 1.75 eV.

The second problem, poor carrier transport, is reduced by introducing an intrinsic layer (which, in practice, is usually slightly n‐type) into the p–n junction giving the p–i–n structure shown in Figure 2.27. This i‐layer greatly increases the width of the depletion region and the associated electric field that sweeps minority carriers across the junction. Assuming the i‐layer is in fact lightly doped n‐type, the highest electric field occurs at the p–i interface and it is therefore best to design the cell so that light enters through a transparent front contact into a very thin, heavily doped p‐type layer. This ensures that most charge carriers are created near the top of the cell and successfully collected.

Unfortunately, the introduction of an i‐layer has its drawbacks. During initial exposure to strong sunlight, absorption by the i‐layer creates additional defects that aid recombination and reduce cell efficiency. The phenomenon, known as the Staebler–Wronski effect, depends on the total number of photons absorbed and therefore on the intensity and duration of the light and the thickness of the i‐layer. Building up over a timescale of months, it results in final or “stabilized” efficiencies significantly lower than the initial values. In the past this has given single‐junction a‐Si(H) cells a rather doubtful reputation. But most PV clouds have a silver lining. In the case of Staebler–Wronski, the initial loss of efficiency can be largely overcome using multi‐junction or stacked cell structures in which light absorption is shared between two or more much thinner i‐layers. Furthermore, by stacking cells with different bandgaps, it is possible to capture a bigger percentage of solar photons and achieve relatively good levels of efficiency and stability, especially in weak or diffuse sunlight.

The basic scheme for one type of triple‐junction cell is shown in Figure 2.28. It depends on the ability of a‐Si to form good alloys with germanium, producing semiconductor material with smaller bandgaps. The top a‐Si “blue cell” is effective at capturing high‐energy blue photons with its bandgap of about 1.75 eV. Next comes the “green cell,” based on amorphous silicon–germanium alloy containing about 15% germanium with a bandgap of around 1.6 eV. And finally the bottom “red cell,” designed to capture low‐energy red and IR photons, uses an alloy with about 50% germanium giving a bandgap of around 1.4 eV. Photons that are not absorbed on the first pass through the cells are returned by the back reflector which may be texturized to encourage light trapping.

The supporting substrate does not have to be flexible, but flexibility offers exciting possibilities during production and also for the user. The production process can be continuous “roll to roll,” the various layers being deposited on an extremely long thin sheet of stainless steel or plastic as it travels between rollers in the manner of a magic carpet (Figure 2.29). This was the dream of solar cell pioneers back in the 1950s! Sheet thickness is typically a small fraction of a millimeter, with sheet lengths up to an amazing kilometer. Individual solar cells are automatically scribed and interconnected as a monolithic circuit. From the user’s point of view, flexibility tends to go hand in hand with lightness and allows easy mounting on curved or awkward surfaces.

The lack of a crystal structure in amorphous silicon ultimately prevents it from matching the efficiency of crystalline silicon, and its commercial production has been phased out. However recent years have seen much R&D effort directed toward a new microcrystalline form of silicon that, like other thin‐film materials, can be deposited in extremely thin layers of about 1 µm onto various substrates including glass. Crystalline silicon’s comparatively poor light absorption means that success depends upon highly effective light trapping to keep incident light within the film. The hope is that microcrystalline silicon will rival wafer technology for ruggedness and electrical stability while at the same time using minimal amounts of cheap and plentiful raw materials, improving efficiency above amorphous products, and greatly reducing costs.

Silicon and germanium may be the best‐known semiconductors, but they are certainly not the only ones. Many compounds incorporating rather unfamiliar chemical elements also display electrical properties midway between insulators and conductors. Some readily absorb solar photons to produce electron–hole pairs and may be doped to make n‐type or p‐type material and deposited as thin layers on a variety of substrates. In other words they are candidates for “second‐generation” thin‐film cells that surpass amorphous silicon’s efficiency and challenge crystalline silicon’s usage of materials and production costs. Of the various possibilities, two materials—CIGS and related compounds and CdTe—have a highly significant presence in the terrestrial PV market and are set to lead PV decisively into a new era.

Not that crystalline silicon cells will be easily displaced. Global production continues apace. Gigawatts of wafer‐based modules are already installed and will generate electricity for many years to come, catching the public eye as ambassadors for PV around the world. However, CdTe thin‐film technology has been over the last 10 years the leader in manufacturing cost reductions and may offer the best chance of achieving grid parity with conventional electricity generation on a large scale.

In the following text, we discuss CIGS and CdTe, and in Section 2.5 we will discuss even more exotic third‐generation technologies.

2.3.2 Copper Indium Gallium Diselenide (CIGS)

To be successful, inorganic crystalline solar cell materials need two essential properties. They must be good light absorbers, turning solar photons into electron–hole pairs; and they must include an efficient junction to sweep light‐generated minority carriers across the junction and force current through an external circuit.

Many years ago it was discovered that the compound semiconductor copper indium selenide (CIS) offers excellent light absorption in small‐grained layers a micrometer or two thick. Although the electronic and chemical properties of CIS and related compounds are subtle and complex,^6,7 a few key points can be made here. First, and unlike silicon, CIS cannot be doped to form an efficient p–n junction on its own (it cannot form a homojunction); but it can be interfaced with another semiconductor, cadmium sulfide (CdS), to produce an effective heterojunction. CIS and CdS are well matched and do not suffer excessive recombination at the interface. Since CdS can only be successfully doped as n‐type material, the CIS must be doped p‐type. It is rather difficult to make good metallic contact with CIS; gold is effective, but expensive, so molybdenum is normally used as a back contact.

There is a further important twist to the story. In the 1970s it was discovered that the rather low bandgap of CIS (about 1.1 eV) may be increased by substituting some gallium in place of indium. By varying the gallium content, a range of bandgaps relevant to PV cells can be obtained, from about 1.1 eV (no gallium) up to 1.7 eV. In addition, the low open‐circuit voltage of CIS is raised toward 0.5 V, comparable with crystalline silicon, meaning that fewer cells need be interconnected to achieve useful module voltages. The modified material, copper indium/gallium diselenide (CIGS), has achieved many cell efficiency records (it is worth noting that the initials CIS and CIGS tend to be used interchangeably, which can lead to a certain amount of confusion). CIGS passed the 20% efficiency milestone for laboratory cells in 2008. At that stage commercial module efficiencies were already attaining 10–12%, comfortably beating amorphous silicon and within aiming distance of crystalline silicon.

The basic scheme of a typical CIGS cell is shown in Figure 2.30. Light enters the cell via a transparent conducting layer acting as the top contact. Next comes an extremely thin layer of CdS that forms a p–n heterojunction with the thicker (but still very thin!) CIGS absorber. A metallic layer, normally molybdenum, provides the back contact and completes the electronic design. The doping of the p‐type absorber is often graded, being lightest near the junction. This extends the depletion region and its associated electric field well into the absorber where most charge carriers are generated and helps sweep them across the junction. Not shown in the figure is the necessary supporting substrate, which may be rigid or flexible and made of glass, metal, or plastic.

As our attention moves from silicon cells with their superabundance of cheap raw material to thin‐film cells based on unfamiliar elements, it is time to question cost and availability of supplies. Cost is not generally seen as a problem, given the tiny amounts of material used in thin‐film cells compared with silicon wafers; indeed one of thin‐film technology’s main promises is to make PV ever more affordable. But the situation could change if production levels continue to increase dramatically. The indium used in CIS and CIGS cells is a case in point: indium is a comparatively rare element of the Earth’s crust, in demand for electronic products other than solar cells. Availability may become a problem in large penetration scenarios and one advantage of partially substituting gallium into CIGS is a decreased demand for indium. This topic is further discussed in Chapter 7.

As thin‐film solar cells contribute more and more to “second‐generation” PV technology and challenge the pole position occupied for so long by crystalline silicon, we will become used to seeing CIS and CIGS modules with the smooth, dark grey/black appearance (shown in Figure 2.31) often favored by architects. There is also intensive development of semitransparent modules that act as windows, allowing a portion of light to enter a building while at the same time generating electricity. The possibilities for exciting and innovative PV products are enormously increased by thin‐film techniques.

2.3.3 Cadmium Telluride (CdTe)

Cadmium telluride (CdTe) is another important semiconductor material for thin‐film solar cells, its direct bandgap of 1.45 eV being close to optimum for capturing the sun’s spectrum using a single‐junction device. Also its high optical absorption coefficient allows light to be fully captured using only a 1.5‐µm‐thick layer. Like many II–VI compounds, CdTe sublimes congruently; it vaporizes homogeneously and the compound’s thermodynamic stability simplifies the deposition of layers of stoichiometric CdTe.⁷ However, initially there was considerable concern among environmental groups about the commercialization of CdTe cells and modules because of cadmium’s toxicity. However, these fears have been largely allayed as the life cycle of CdTe PV has been critically reviewed by expert panels in more than 12 countries and all concluded that the technology is safe and friendly to the environment.^8–10 CdTe has not got the toxicity of its individual constituents Cd and Te. Cadmium is commonly obtained as a byproduct of zinc mining and smelting, so removing it from the environment for use in solar cells offers an environmental benefit, especially when modules are recycled at the end of their useful life. This issue is discussed in Chapter 7. Cadmium and tellurium are more abundant elements than the indium used in CIS/CIGS products, so availability is not so big an issue. However, the market is growing strongly. CdTe modules accounted for about 10% of world production in 2016, a lot more than any other thin‐film technology, and they are finding large‐scale application in PV power plants. Comparatively simple production processes mean that CdTe modules are currently about the cheapest on the market in terms of price per peak watt. Furthermore their conversion efficiencies of around 16.5% (2016 average) look set to advance toward 20% in the next few years.

The rationale behind a thin‐film CdTe solar cell results in a scheme very similar to that for CIS and CIGS. The essential layers in the thin‐film “sandwich” are a transparent top contact, a CdS/CdTe p–n heterojunction and absorber, and a metallic back contact, as shown in Figure 2.32. Also required is a suitable supporting substrate of glass, metal, or plastic, which determines whether cells are rigid or flexible. Bear in mind that although the figure represents the cell as rather thick and narrow, it is actually manufactured as part of an extremely thin sheet.

As worldwide thin‐film PV production grows, and it seems that cadmium telluride will continue its important contribution. An installation that nicely illustrates the possibilities for “farming sunshine” alongside conventional crops is shown in Figure 2.33. Further up the power scale, two 550 MW_ac power plants with CdTe modules made by First Solar have been commissioned in south California (Figure 2.34); such sizes were almost unimaginable 10 years ago.

2.5.1 Gallium Arsenide (GaAs) Multi‐Junctions

Gallium is one of the elements in Group III of the periodic table; arsenic is in Group V. So gallium arsenide (GaAs) is often referred to as a Group III–V semiconductor. GaAs and associated compounds have two claims on our attention as specialized, but important, PV materials: for making solar cells used in spacecraft and for their use in terrestrial concentrator systems that focus sunlight using mirrors or lenses.

In the early years of space exploration, silicon solar cells were the main source of electricity for spacecraft, reaching efficiencies of about 15% by 1970. Since then GaAs has made a big impact, for two main reasons. First, it is less susceptible than silicon to damage by radiation in space, a key consideration on long missions where the performance and reliability of electricity supply is paramount. Second, its direct bandgap of 1.42 eV (compared with 1.1 eV for silicon) allows a greater percentage of the solar spectrum to be harvested, giving better conversion efficiencies. Since the 1980s solar cell designers have learned how to deposit thin films on crystalline germanium wafers, producing lightweight multi‐junction devices of even higher efficiency. Triple‐junction modules have gained a high reputation for their reliability and light weight. And although the material and processing costs of GaAs cells are high, this is hardly a major consideration for vastly expensive space projects.

A typical scheme for triple‐junction GaAs cells is shown in Figure 2.38. Like the triple‐junction amorphous silicon cell described earlier, it is a “sandwich” of three stacked cells with different bandgaps designed to capture different portions of the sun’s spectral energy. For space applications the relevant spectrum corresponds to air mass zero (AM0), received by solar cells outside the Earth’s atmosphere (refer back to Figure 1.8). Each cell includes n‐type and p‐type crystalline layers. The top cell, with a bandgap of about 1.9 eV obtained using the alloy gallium indium phosphide (GaInP), is very effective at absorbing high‐energy UV/blue photons. The GaAs cell in the middle has a bandgap of 1.42 eV; and the bottom cell, based on germanium that also provides the supporting substrate, has a bandgap of 0.7 eV to absorb IR photons.

Although such triple‐junction devices come in the general category of “gallium arsenide,” we see that they actually use carefully controlled proportions of several III–V elements plus Group IV germanium to achieve bandgap control. These highly specialized solar cells are built up monolithically, with many layers being grown on top of one another with optimal thickness and doping. All this requires expensive materials and very advanced processing. But the technical rewards are high: the best laboratory cells have efficiencies of 40% and commercial cells over 35%. Such impressive efficiencies pose an interesting question. Can gallium arsenide be “brought down to Earth” and make a significant contribution to terrestrial PV generation? Success depends upon effective concentration of sunlight using mirrors or lenses, focusing the light onto cells of far smaller area with correspondingly reduced material and processing costs. For example, increasing the light intensity 1000 times (“1000 Suns”) should allow the cell area to be reduced 1000 times for the same power output. Indeed, it is rather better than this because the efficiency of many solar cells improves under concentrated sunlight. Triple‐junction GaAs concentrator cells have already passed the 44% landmark in the laboratory, with commercial cells not far behind.

Successful PV concentration systems must aim to reduce cell costs sufficiently to offset the expense of focusing the light and tracking the sun across the sky on its daily journey. Not surprisingly, there are sceptics; yet PV concentration is being intensively researched and developed, with many systems in commercial production.

2.5.2 Dye‐Sensitized Cells

Some of the new PV concepts and materials introduced in recent years would have astounded early PV pioneers whose attention was entirely focused on inorganic semiconductors, principally silicon and germanium. We are now moving into an era where artificial organic materials seem certain to play an important role in converting sunlight directly to electricity. They are seen as part of PV’s “third generation.” Of many possible approaches, Dye‐Sensitized Cells (DSCs) and organic cells are presently in the vanguard of development and commercial application.

DSCs are a hybrid organic–inorganic technology that uses small‐molecule absorber dyes absorbed onto an electron‐accepting material, such as titanium dioxide (TiO₂), along with an electrolyte to regenerate the dye. They are also called the “Graetzel cells” after Michael Graetzel who with Brian O’Regan at the Federal Polytechnic in Lausanne, Switzerland, found that a 10 µm thin film of TiO₂ could work as an effective solar cell if coated with an organic dye, immersed in an electrolyte, and provided with electrical contacts. Most importantly, the TiO₂ was made in the form of a nanoporous “sponge” of minute particles just tens of nanometers (nm) across, propelling PV into the modern field of nanotechnology. And since titanium dioxide (also known as titania) is an inorganic semiconductor, whereas the dye and electrolyte are organic, the Graetzel cell is sometimes referred to as an organic–inorganic thin‐film device.

But why dye sensitized? Unlike conventional cells in which the absorption of light and transport of light‐generated charges takes place within the same semiconductor, in a DSC these roles are split. The dye acts as light absorber, generating electrons that it injects into the conduction band of the semiconductor. In other words the dye acts as a “sensitizer” of the TiO₂, which would not be effective on its own because of its large bandgap. Another key aspect of the Graetzel cells is their use of new organic dyes able to absorb a wide solar spectrum. And the use of TiO₂ nanoparticles, rather than larger crystals, hugely increases the surface area of the adsorbed dye coating and hence the efficiency of light absorption.

Most people, meeting DSCs for the first time, find their detailed operation very complex—certainly more so than crystalline silicon cells. Although it involves many of the same basic concepts¹²—photon absorption, charge generation and transport, recombination, optical and resistance losses—the electrochemical terminology is unfamiliar and the names of the organic materials can seem unreasonably long! So we restrict ourselves here to a brief summary.

The basic scheme of a DSC is illustrated in Figure 2.39. Light enters the cell via a transparent front contact and is absorbed by the organic dye covering the TiO₂. Excitation electrons are injected into the conduction band of the TiO₂, causing oxidation of the dye. They are efficiently transported through the semiconductor by diffusion and reach the electrical contact. Assuming the cell is connected to an electrical load, the electrons now pass through the external circuit and reenter the cell via the back contact or counter electrode. Here they provide the negative charges required to restore the dye to its original (unoxidized) state with the help of the intervening electrolyte. The circuit is completed.

Unfortunately some recombination does occur, but not in the same manner as in conventional silicon cells. Although electrons are injected by the dye into the conduction band of the semiconductor, holes are not formed in its valence band; so there is no generation of electron–hole pairs, or subsequent annihilation. But electrons can recombine with the oxidized dye. Fortunately, electron injection and transport in the semiconductor is extremely fast compared with the recombination process, so effective charge separation does in fact take place. Overall, the photon‐to‐electron generation process in a DSC is analogous to photosynthesis in leaves and plants where chlorophyll acts as the sensitizer.

Early Graetzel cells achieved very respectable efficiencies of up to about 10% in standard insolation conditions (1000 W/m², 25°C). A great deal of ongoing research has since improved performance, raising efficiency above amorphous silicon and within sight of other thin‐film technologies. However, efficiency in bright sunshine is probably not the main criterion for DSCs. They work well in low diffused light and in high ambient temperatures, indoors and out. Flexible modules can easily be made using plastic substrates. They use nontoxic and plentiful materials (TiO₂ is a widely used chemical, e.g., in paints and toothpastes) and their relatively simple manufacturing techniques include fast roll‐to‐roll production. Unusual and exciting possibilities are opening up for building‐integrated photovoltaics (BIPV), including roofing products, transparent and semitransparent tinted windows, partitions, and decorative features. Instead of restricting architects to standard rectangular PV modules, DSC products can be tailor made to particular sizes, shapes, and aesthetic design criteria; see, for example, Figure 2.40. The wide range of applications promises an exciting future.

2.5.3 Organic Solar Cells

In addition to the inorganic semiconductors we examined before, there are organic compounds that can absorb photons or carry electric charges, functioning as semiconductors. These are oligomers or polymers made up by carbon and hydrogen atoms with, sometimes, added nitrogen, sulfur, and oxygen. Like silicon, in general they are insulators but become semiconducting when doped, or by photoexcitation. OPV cells use long‐chained molecular systems for the electron‐donating material (e.g., P3HT), along with fullerenes as the electron‐accepting system (e.g., PC60BM); see the box in the next page for a basic description of fullerene structure.

They attracted interest for PV applications because they use inexpensive materials and can be fabricated with low‐cost, solution‐based processing. Soluble organic molecules enable roll‐to‐roll processing techniques and allow for low‐cost manufacturing. Also these materials can be applied to flexible substrates enabling a wide variety of uses.

Progress in the development of OPV has been fast; as shown in Figure 2.35, OPV record cell efficiencies have gone from 3 to 11% within 14 years of development, and tandem cells that have the potential for exceeding the Shockley–Queisser limit reached the same record efficiency within only 6 years (2008–2015). They have been commercialized, amid small volumes, by Konarka and Heliatek since the potential for new applications counterbalances their low efficiency. The area where OPV is expected to have the largest impact is in BIPV where they offer shape and design flexibility since they can be tuned to the desirable colors by slightly changing their chemical properties, thereby allowing solar cells to be an integral part of the design. Currently only Heliatek produces OPV; these are used for retrofits on window and building exteriors. Their vision is that eventually OPV could be used to create solar‐coated cars and homes.

Let’s now see how OPV functions. Polymer‐based OPV cells use long‐chained molecular systems for the electron‐donating material (e.g., poly 3‐hexylthiophene (P3HT)), along with fullerenes as the electron‐accepting system (e.g., C60PCBM). The absorber is used in conjunction with an electron acceptor, such as a fullerene, which has molecular orbital energy states that facilitate electron transfer. Photon absorption in OPV does not lead directly to an electron and a hole, but it first generates an exciton, a state where the two charges are bound together. The exciton then migrates to the interface (heterojunction) between the absorber material and the electron acceptor material. At the interface, the energetic mismatch of the molecular orbitals provides sufficient driving force to split the exciton and create free‐charge carriers (an electron and a hole).

The most common device structure for OPV uses a mixture of donor and acceptor materials referred to as a bulk heterojunction (BHJ) that resides between two electrodes. Figure 2.41 (a) illustrates a BHJ (green‐blue colors) packed between an electron blocking layer (EBL) and a hole blocking layer (HBL), which are in contact with an indium tin oxide (ITO) and a silver electrode.14 As depicted, the PV effect follows the following steps: the illumination of an organic semiconductor donor (1) generates excitons (2) with a binding energy of about 0.4 eV. To separate into free charges, the exciton must diffuse until it reaches a donor/acceptor interface (3) with a difference in electron affinities and an ionization potential large enough to overcome the binding energy. The energy cascade required for charge extraction is illustrated in Figure 2.41(b). The free charges then can travel (4) through either the donor or acceptor material (5), and then are collected at the electrodes (6). The overall efficiency of the device therefore is determined by the optical absorption and the efficiency of each of those steps.

Various challenges have limited the use of OPV; in particular, the large bandgap of most organic polymers is responsible for low power‐conversion efficiency because a large portion of the solar spectrum is unabsorbed. In theory, in an optimal solar cell, the acceptor bandgap should be around 1.4 eV, wherein the maximum efficiency would be 31% under 1 sun AM1.5. Most early‐generation semiconducting polymers have bandgaps higher than 2 eV (corresponding to a wavelength of 620 nm), so limiting their maximum efficiency. There are two alternatives to increase the devices’ efficiency: lowering the bandgap to absorb a maximum of photons in one layer or using a multi‐junction approach where two different materials absorb in a different region of the solar spectrum.

The low bandgap approach has received considerable interest in the last few years and produced the current record efficiency for polymer devices. By lowering the bandgap from 2 to 1.5 eV, the maximum theoretical efficiency increases from 8 to 13%. To increase efficiencies further, the multi‐junction approach illustrated in Figure 2.42 is necessary. For OPV this approach not only is advantageous to capture a broader range of the solar spectrum but also helps overcome the poor charge carrier mobility and lifetime of carriers, which prevents the fabrication of a thick absorbing layer. In comparison with inorganic material, organic semiconductors absorb only a narrow portion of the spectrum and therefore a combination of multiple materials is necessary to enhance photon absorption.

Also note that because various absorbers can be used to create colored or transparent OPV devices, this technology is particularly appealing to the BIPV market. OPV has achieved efficiencies near 11%, but efficiency limitations as well as long‐term reliability remain significant barriers; these challenges present opportunities for further research, which we discuss in the following text.

2.5.4 Perovskites

Perovskites are the latest addition to the promise of third‐generation PV. Figure 2.35 shows the extremely rapid efficiency improvements in perovskite cells. In just 5 years, lead‐based organometal halide perovskite solar cells have shown in the lab efficiencies up to 22%. This matches the best efficiencies of CdTe and CIGS thin‐film device technologies that have been developed over several decades. However, high efficiencies were observed only at very small (i.e., 1 cm² or less) sizes that become unstable within minutes, whereas PV is supposed to last 30 years. Perovskites have the crystallographic structure ABX3, where A is a large cation such as methyl‐ammonium, B is typically lead (Pb), and X is halogen or a mixture of halogens (I, Br, and Cl). They are appealing because they have large absorption coefficients and offer a range of bandgaps for both single‐junction devices and Si‐based tandem cells. Moreover, perovskites are less sensitive than other technologies to structure defects and show long carrier lifetimes and diffusion lengths. The rapid improvement in device performance has been largely based upon improved understanding and control of composition, microstructure, and complementary charge transport layers in the device structure. Early reports of rapid improvements in efficiency led to an explosion of research efforts, with well over 1000 publications on the topic in 2015.

Despite great excitement, commercialization of perovskite PV technology requires overcoming three critical challenges: demonstrating long‐term operational stability of PV modules, scaling up to large areas, and avoiding or mitigating real and perceived risks associated with the use of toxic lead. Regarding stability and size, early lab‐scale perovskite solar cells have exhibited significant degradation of performance on timescales of minutes although their size is as small as a fingernail. The causes of this degradation are only partially understood. Proper encapsulation will address some of the stability issues related to moisture exposure, but work to modify the device layers themselves has also shown promise.

The other challenge to commercialization is the use of toxic lead. The intensity and risks associated with lead in perovskite solar cells are subjects of controversy. Some reports have indicated that any amount of lead is unacceptable, whereas other reported that the amount of lead potentially used in PV would be negligibly small compared with lead usage in from other industries such as in electronic solder, aviation fuel, and coal‐fired electricity production.