Review of CdTe_1−xSe_x Thin Films in Solar Cell Applications

Abstract

Recent improvements in CdTe thin film solar cells have been achieved by using CdTe_1−xSe_x as a part of the absorber layer. This review summarizes the published literature concerning the material properties of CdTe_1−xSe_x and its application in current thin film CdTe photovoltaics. One of the important properties of CdTe_1−xSe_x is its band gap bowing, which facilitates a lowering of the CdTe band gap towards the optimum band gap for highest theoretical efficiency. In practice, a CdTe_1−xSe_x gradient is introduced to the front of CdTe, which induces a band gap gradient and allows for the fabrication of solar cells with enhanced short-circuit current while maintaining a high open-circuit voltage. In some device structures, the addition of CdTe_1−xSe_x also allows for a reduction in CdS thickness or its complete elimination, reducing parasitic absorption of low wavelength photons.

1. Introduction

With a global drive towards renewable energies, photovoltaics is an increasing part of global energy supply. CdTe is the leading technology amongst thin film solar cells for cost effective solar electricity production, due to its high photovoltaic conversion efficiency, long-term performance stability, low fabrication costs and short energy-payback time [1]. The current efficiency record for CdTe cells is 22.1%, held by First Solar [2,3].

In state-of-the-art CdTe solar cells, the J_SC is already close to its theoretical limit [4]. According to the Shockley–Queisser thermodynamic limit to solar cell efficiency, the optimum band gap for maximum theoretical solar cell efficiency under AM 1.5 G is 1.34 eV [5,6]; the band gap of CdTe is slightly too wide at 1.5 eV [4]. Strategies to shift the band gap of the CdTe absorber towards the theoretical optimum can help to further increase photocurrent and, consequently, the efficiency. CdTe_1−xSe_x is a promising material for this purpose, as its band gap can be shifted down to 1.4 eV [7,8] by appropriately adjusting the composition. First Solar’s world record cell efficiency was improved from 19.5% to 22.1% using an absorber with a lower band gap than CdTe, the CdTe_1−xSe_x alloy [9,10,11].

With an absorber material that has an adjustable band gap, the band gap profile through the absorber layer can be engineered by applying a compositional grading. Different research groups have been investigating the use of a CdTe_1−xSe_x layer towards the front of a CdTe absorber layer, resulting in champion devices with improved J_SC due to higher absorption in the low-band gap front layer, while maintaining the V_OC from the bulk CdTe [8,12,13,14,15,16,17].

For traditional CdS/CdTe solar cells, a CdCl₂ treatment, generally consisting of deposition of CdCl₂ on the absorber and subsequent heating, is needed to recrystallize the absorber layer, resulting in increased CdTe grain size and intermixing of CdS and CdTe as well as modifications in electronic properties leading to enhancement in photovoltaic conversion efficiency [18]. If a CdSe or CdTe_1−xSe_x layer is inserted in between CdS and CdTe, interdiffusion during CdCl₂ treatment results in a CdTe_1−xSe_x layer. Because CdTe_1−xSe_x has no miscibility gap, a continuous composition gradient is formed, with an associated band gap gradient [15,19], allowing for absorption of high-wavelength photons. Several research groups have also found that, with the use of a CdTe_1−xSe_x front layer, the CdS window layer thickness can be reduced, or the CdS layer can even be omitted completely, reducing parasitic absorption in the short-wavelength region and further increasing J_SC [8,14,20]. A thick CdS layer is generally required to avoid direct contact between the transparent conducting oxide (TCO) and the CdTe absorber, but a CdTe_1−xSe_x absorber seems to be less sensitive to this issue. Further improvement was achieved by replacing CdS with MgZnO, which has a more favorable band alignment with both CdTe and CdTe_1−xSe_x than CdS, and which is more transparent in the short-wavelength region [15,21,22]. Additionally, passivation of defects due to Se has been reported to be responsible for the higher carrier lifetimes in CdTe_1−xSe_x compared to CdTe [19,23].

CdTe technology has shifted in recent years from a traditional CdS/CdTe structure to a device structure that utilizes a CdTe_1−xSe_x layer and a band gap gradient, but no CdS layer, in order to achieve high currents and voltages. In this review paper we discuss the steps of this transformation, from the first introduction of a CdTe_1−xSe_x layer to the complete replacement of the CdS layer by MgZnO and CdTe_1−xSe_x. We discuss the material properties of CdTe_1−xSe_x and their dependence on composition and the application of this material in thin-film solar cells.

2. Growth of CdTe_1−xSe_x Thin Films and Material Properties

In this section, we will discuss the deposition of CdTe_1−xSe_x films and their structural, optical and electronic properties.

CdTe_1−xSe_x is a solid solution without a miscibility gap; the alloy is stable over the complete composition range 0 ≤ x ≤ 1. Thin films of desired composition can be grown with a variety of methods, such as high-vacuum evaporation of the elements [24] or the alloy CdTe_1−xSe_x [25,26], co-evaporation of CdTe and CdSe [8,27], close space sublimation of the alloy CdTe_1−xSe_x [23,28] or co-sublimation of the components CdTe and CdSe [29], molecular beam epitaxy [30,31], hot wall deposition [32], electron beam deposition [7,33], or electrodeposition [34,35]. The properties of grown layers depend on the deposition method and the conditions used.

In solar cells, the main methods used for CdTe_1−xSe_x formation are high-vacuum evaporation [8] or radio frequency sputtering [14,16,20] of a CdSe layer next to a CdTe layer, with interdiffusion during CdCl₂ treatment, or close space sublimation of a CdTe_1−xSe_x layer next to a CdTe layer [15], which also interdiffuse during CdCl₂ treatment. Close space sublimation is performed at higher temperatures than high-vacuum deposition or sputtering, and in CdTe it results in very large grains. The lower deposition temperature of high-vacuum evaporation results in smaller as-deposited grains. However, recrystallization during CdCl₂ treatment results in overall comparable grain sizes between different deposition methods [36]. The properties of CdTe_1−xSe_x are therefore expected to depend on deposition parameters as well as CdCl₂ treatment conditions.

2.1. Crystal Structure

CdTe crystallizes in the cubic zinc blende structure, while CdSe crystallizes in the hexagonal wurzite structure. The two structures are closely related; each Cd²⁺ ion is surrounded by a tetrahedron of four Se²⁻ and/or Te²⁻ ions. In the solid solution CdTe_1−xSe_x, both the zinc blende and the wurzite structure have been reported [16,34,37,38,39]. For 0.2 ≤ x ≤ 0.5, cubic zinc blende is the low-temperature structure, with a transition to hexagonal wurzite at temperatures above 800 °C. This transition temperature decreases with increasing x, and at x = 0.6 the hexagonal wurzite structure is formed at 600 °C [16,37]. With appropriate deposition parameters, mainly dependent on substrate temperature, the entire composition range of CdTe_1−xSe_x can be fabricated in the zinc blende structure. In this structure, the lattice constant a decreases with increasing x, following Vegard’s law, due to the smaller size of Se²⁻ compared to Te²⁻ [7,33,40,41]. The wurzite structure can occur at room temperature for x = 0.3 and higher, sometimes mixed with the zinc blende structure [34,38,39]. In solar cell applications, the zinc blende structure is desired, as it is photoactive, i.e., it can convert light into photovoltaic current, while the wurzite structure is not photoactive [16].

2.2. Optical Properties

CdTe_1−xSe_x has a direct band gap with a pronounced bowing behavior. Figure 1 shows reported band gaps of CdTe_1−xSe_x thin films measured with different methods, as well as one band gap bowing derived from first-principles DFT calculations, all showing a similar bowing behavior trend. The bowing parameter b can be extracted from experimental data with a second-degree polynomial fit. The band gap of any composition x can then be calculated using Equation (1) [42]:

$$E_G\left(x\right)=x\cdot E_G^{CdSe}+\left(1-x\right)\cdot E_G^{CdTe}-b\cdot x\cdot \left(1-x\right)$$

(1)

Band gap bowing parameters from bulk and thin-film CdTe_1−xSe_x, fabricated by different methods and using different measurement techniques, are listed in

Table 1. Band gap bowing parameters of layers and powders synthesized with different methods.

Source	Deposition	Band Gap Determination	Crystal Structure	x Range	Bowing Parameter	x of Band Gap Minimum	Band Gap Minimum
[39]	Sintering of mixed CdTe and CdSe powders	Transmittance/reflectance	zinc blende	0–0.6	0.74	0.35	1.38
[39]	Sintering of mixed CdTe and CdSe powders	Transmittance/reflectance	wurzite	0.4–1	0.68	0.33	1.4
[8]	Co-evaporation of CdTe and CdSe	Transmittance/reflectance	zinc blende	0–0.5	0.78	0.37	1.39
[7]	Electron-beam evaporation	Transmittance/reflectance	zinc blende	0–1	0.567	0.37	1.44
[46]	Thermal evaporation of mixed CdTe and CdSe powders	Transmittance/reflectance	zinc blende	0–1	0.571	0.35	1.44
[31]	Molecular beam epitaxy on Si	Photoluminescence	zinc blende	0–1	0.97	0.40	1.35
[31]	Molecular beam epitaxy on Si	Cathodoluminescence	zinc blende	0–1	0.82	0.38	1.39
[42]	–	First-principles DFT calculations	zinc blende	0–1	0.75	0.40	1.41
[43]	–	sp3s* tight-binding method calculations	zinc blende	0–1	0.904	0.39	1.37
[44]	–	First-principle hybrid-functional calculations	zinc blende	0–1	0.725	0.39	1.43

. The published bowing parameter values cover a range of 0.56–0.97, and they all agree on the composition of the band gap minimum at 0.3 ≤ x ≤ 0.4, independent of fabrication method. The bowing only marginally depends on the crystal structure [39]. The band gap determination using photoluminescence and cathodoluminescence seem to yield bowing parameters slightly higher than transmittance/reflectance measurements, but they are still in good agreement [31].

Theoretical calculations using different methods are also shown in Table 1, and they seem to be in good agreement with experimental data [42,43,44]. The band gap minima in all experiments and calculations are around 1.4 eV, confirming that CdTe_1−xSe_x is a good candidate for band-gap engineering in CdTe solar cells for the purpose of increasing J_SC.

In CdTe_1−xSe_x, the valence band edge is dominated by Se and Te p-orbitals, and the conduction band edge is dominated by Cd s- and p-states. Therefore, the anion substitution would be expected to primarily affect the valence band edge [43,45]. However, first-principles calculations show that the bowing of the band gap is caused by the bowing of both band edges due to strong intra-band coupling in both the valence and the conduction band [44].

The refractive index of CdTe_1−xSe_x is almost constant over a large range of wavelengths, as well as with changes in composition and film thickness. Use of a CdTe_1−xSe_x layer is therefore not expected to change the overall reflectance of a solar cell, and the introduction of a CdTe_1−xSe_x gradient does not introduce a gradient in the refractive index [7,25,32].

2.3. Electronic Properties

The electronic properties of CdTe_1−xSe_x have not been studied extensively. As-deposited CdTe_1−xSe_x is an n-type semiconductor with carrier densities between 10¹³ cm⁻³ (CdTe) and 10¹⁸ cm⁻³ (CdSe) [40]. The Hall mobility increases by less than an order of magnitude between CdTe and CdSe. Both carrier density and mobility are dependent on deposition temperature, with higher values at higher deposition temperatures due to increased grain size and possibly a reduction in structural defects [26,40]. In solar cell processing, the CdTe_1−xSe_x layer is subjected to a CdCl₂ treatment, during which it recrystallizes. In CdTe, the recrystallization results in larger grains, reduced structural defects, and reduced recombination [18,47]. A similar effect is expected for CdTe_1−xSe_x. In absorbers consisting of a CdTe_1−xSe_x layer in front of CdTe, recrystallization during CdCl₂ treatment has been shown to be responsible for a reduction in stacking faults, combined with interdiffusion of the CdTe_1−xSe_x and CdTe layers [19,28].

In solar cells that use CdTe_1−xSe_x as part of the absorber layer, dopants are generally added to the whole CdTe_1−xSe_x/CdTe stack and permeate the CdTe_1−xSe_x layer via diffusion. A commonly used extrinsic dopant is copper, which forms deep acceptors as Cu on a Cd site (Cu_Cd) [48,49]. First-principles calculations indicate that the formation energy of this acceptor defect exhibits a bowing, with lower formation energies in the CdTe_1−xSe_x solid solution than in either parent compound [44]. In the mixed alloy, the bonding Te orbitals are partially replaced with Se orbitals, which have a lower energy. The electronic effects of this change in bonding orbitals are one reason for the bowing behavior. Another reason is the smaller lattice constant with increasing x leading to reduced compressive strain when Cd is replaced by the much smaller Cu [44]. The reduced formation energy indicates better dopability of CdTe_1−xSe_x with copper compared to CdTe. However, Hall effect measurements of CdTe_1−xSe_x with 0 ≤ x ≤ 0.2, deposited by high-vacuum evaporation and subjected to a CdCl₂ treatment, revealed that the achievable doping concentration is lower in Se-containing samples, with more than an order of magnitude difference between CdTe and x = 0.2 [8]. A possible explanation is that compensating donors, such as copper on interstitial sites (Cu_i), are formed as well, and that the formation energy of these defects is also dependent on x. If the electronic and/or structural effects of CdTe_1−xSe_x alloying facilitate the formation of compensating donors more than the formation of acceptor defects, the result is a limit to the achievable hole density [8,49,50].

The photoluminescence decay of CdTe_1−xSe_x measured by time-resolved photoluminescence (TRPL) is generally dominated by surface recombination, resulting in short apparent lifetimes [23,51]. In order to get a reliable measure for the bulk minority carrier lifetime, the surface has to be passivated, e.g., using Al₂O₃, which provides field-effect passivation [52]. Another approach is two-photon excitation (2PE) TRPL, which allows for lifetime measurements at a selected depth of the absorber [51,53]. In CdTe_1−xSe_x layers (x = 0.2, deposited by close space sublimation and subjected to CdCl₂ treatment) with both surfaces passivated by sputtered Al₂O₃, lifetimes of up to 430 ns were measured, an order of magnitude higher than lifetimes in similarly passivated CdTe [23]. The reason seems to be that Se passivates a deep defect in the bulk of the absorber, even at low concentrations [19]. This is also correlated with longer diffusion length in material with higher x [13,19]. Two-photon excitation TRPL measurements performed on an interdiffused CdTe_1−xSe_x/CdTe absorber (close space sublimation, CdCl₂ treated, resulting maximum Se content x ≈ 0.16) have shown lifetimes of around an order of magnitude higher in the Se-rich front layer compared to the CdTe back layer. This lifetime difference within a single absorber was linked to reduced recombination at the grain boundaries in the Se-containing region [51,53]. Se therefore, at least up to x ≈ 0.2, seems to provide both passivation of a bulk defect and of the grain boundaries, resulting in an overall increased lifetime. Further studies are needed to accurately determine the lifetime dependence with x, and to determine the steepness of the lifetime gradient in an interdiffused CdTe_1−xSe_x/CdTe absorber.

The various deposition methods used for CdSe or CdTe_1−xSe_x apply a variety of substrate temperatures and deposition rates. We expect that the reported properties of CdTe_1−xSe_x may depend on deposition method as well as on the parameters of the CdCl₂ treatment, but further investigations are needed to determine the extent.

3. CdTe_1−xSe_x in Solar Cells

In recent years, CdTe_1−xSe_x has been used as a layer at the front of a predominantly CdTe absorber layer (see Figure 2), either deposited as CdSe or as CdTe_1−xSe_x, which interdiffuse with CdTe during CdCl₂ treatment to form a continuous composition gradient. This results in an absorber layer with a low band gap, increasing the J_SC, and it allows for a reduction in CdS thickness, both in superstrate and substrate configuration. As a further step, Munshi et al. [15] replaced the CdS with MgZnO for the highest device efficiencies. Schematics of the different device configurations in which CdTe_1−xSe_x layers have been used are shown in Figure 2, from CdS/CdTe_1−xSe_x/CdTe (Figure 2b), to CdTe_1−xSe_x/CdTe (Figure 2c), to MgZnO/CdTe_1−xSe_x/CdTe (Figure 2d). As an alternative to the conventional superstrate configuration, a device in substrate configuration (Figure 2e) is discussed as well.

The best reported devices for each of the different structures in Figure 2 are listed in

Table 2. J-V parameters of the best CdTe_1−xSe_x-containing devices for each configuration shown in Figure 2.

Source	Device Structure	VOC [mV]	JSC [mA/cm2]	FF [%]	Efficiency [%]
Paudel and Yan [20]	Figure 2b	806	27.2	64.1	14.1
Paudel and Yan [20]	Figure 2c	771	27.5	69.4	14.7
Munshi et al. [15]	Figure 2d	854	28.4	79.1	19.1
Lingg et al. [8]	Figure 2e	710	25.6	67	12.2
First Solar [3,9,10]	unknown	887.2	31.69	78.5	22.1

, as well as the CdTe_1−xSe_x-containing record device by First Solar. By removing the CdS layer, a small increase in J_SC and FF has been shown, due to reduced parasitic absorption of CdS, at the cost of some losses in V_OC, due to interface recombination at the TCO/CdTe_1−xSe_x interface [20]. With a MgZnO window layer, all device parameters could be improved: high J_SC due to the absence of parasitic absorption in the window layer, and high V_OC and FF due to a good band alignment between MgZnO and CdTe_1−xSe_x, resulting in a power conversion efficiency of 19.1% [15]. For the substrate configuration device, while the FF is comparable to superstrate devices, the V_OC and J_SC are lower. The loss in V_OC may be due to interface recombination at the CdS/CdTe_1−xSe_x interface; the loss in J_SC due to parasitic absorption in the remaining CdS layer [8]. The world record device from First Solar surpasses all other devices in terms of V_OC and J_SC, and only the MgZnO/CdTe_1−xSe_x device can exceed its FF. The exact structure and therefore the reasons for this outstanding performance are not public knowledge, but the high J_SC and V_OC indicate the use of MgZnO or a similar window layer [3,9,10].

In the following section, we will first discuss the deposition methods used to form a CdTe_1−xSe_x/CdTe solar cell absorber, and in subsequent subsections we will discuss the details of how CdTe_1−xSe_x has been used to improve performance in devices with different configurations.

3.1. Formation of a CdTe_1−xSe_x Layer during Device Processing

Different deposition methods have been used to form a CdTe_1−xSe_x layer. In conventional superstrate configuration, CdSe can be deposited by high-vacuum evaporation [27], magnetron sputtering [14,16,47], or pulsed laser deposition [54], with CdTe_1−xSe_x formation during CdCl₂ treatment. An alternative is close-space sublimation (CSS) of a CdTe_1−xSe_x layer [15,19,29]. The CdSe or CdTe_1−xSe_x layer is followed by deposition of CdTe, CdCl₂ and doping treatments, and back contact deposition.

For devices in substrate configuration, a CdSe layer is deposited onto CdTe, followed by CdCl₂ and doping treatments and window layer/front contact deposition (see Figure 2e) [8].

In all of these device structures, the absorber and window layer are recrystallized and the CdSe/CdTe_1−xSe_x interdiffuse with the CdTe during CdCl₂ treatment. In an as-deposited CdTe_1−xSe_x and CdTe absorber, a large number of small grains towards the front (CdTe_1−xSe_x) and increasing grain size towards the back are observed, but after the CdCl₂ treatment, the grains are uniform in size and no distinct boundary is visible between CdTe_1−xSe_x and CdTe [28]. Scanning transmission electron microscopy (STEM)/energy-dispersive x-ray spectroscopy (EDS) mapping reveals interdiffusion of CdTe_1−xSe_x and CdTe, with Se diffusing more than a µm into the CdTe layer [15,28]. Secondary ion mass spectrometry (SIMS) reveals accumulation of Se at grain boundaries in the CdTe layer, and a relative loss of Se at grain boundaries in the CdTe_1−xSe_x layer, which is a strong indication for a combination of fast diffusion of Se along the grain boundaries and slow diffusion into the grains [19].

Any photons with energies above the CdTe band gap (≥1.5 eV) are absorbed in the CdTe_1−xSe_x layer or in the CdTe absorber underneath. CdTe_1−xSe_x has a lower band gap than CdTe (down to 1.4 eV), and it therefore absorbs photons that cannot be absorbed by CdTe. To maximize J_SC gain, the CdTe_1−xSe_x layer has to be sufficiently thick to absorb the majority of these photons with energies below the CdTe band gap. If it is formed from CdSe, the composition of the CdTe_1−xSe_x layer after interdiffusion depends on the thickness of the deposited CdSe layer as well as on the annealing conditions. Yan et al. found that for CdSe layers up to 100 nm thick, the interdiffusion is sufficient to result in a completely photoactive zinc blende absorber with compositions of x ≤ 0.65. For thicker layers, a non-photoactive wurzite layer with x > 0.65 remains at the front of the absorber, which is undesirable because of parasitic absorption. With this method, it is difficult to achieve a thick photoactive layer for full absorption without formation of a non-photoactive layer [16,20,47]. Forming the CdTe_1−xSe_x layer by depositing CdTe_1−xSe_x with x ≤ 0.65 instead of CdSe allows for a higher overall Se content in the absorber because the risk of forming a non-photoactive wurzite layer is minimized, and thick photoactive low-band gap layers can be formed [15].

The interdiffusion of CdTe_1−xSe_x and CdTe creates a compositional gradient, which is expected to be correlated to the gradient in the material properties of CdTe_1−xSe_x, such as band gap, lattice constant, charge carrier concentration, and minority carrier lifetime. In particular, Fiducia et al. [19] reported gradients in both band gap and diffusion length (determined from cathodoluminescence measurements) that are directly correlated with the Se concentration gradient in a CdTe_1−xSe_x/CdTe device.

3.2. Overview of Device Performances with CdTe_1−xSe_x

Different research groups have had varying success in improving device performance using CdSe. Published device parameters are listed in

Table 3. J-V parameters of devices containing CdSe compared with non-CdSe devices. Deposition methods are abbreviated as RFS (radio frequency sputtering), CSS (close space sublimation), and HVE (high vacuum evaporation).

Source	Device Structure	VOC [mV]	JSC [mA/cm2]	FF [%]	Efficiency [%]
Poplawsky et al. [16]	CdSe (100 nm, RFS)/CdTe (CSS)	770	27	60.2	12.6
	CdS (130 nm, RFS)/CdTe (CSS)	810	23.8	75.4	14.5
	difference with CdSe	−40	+3.2	−15.2	−1.9
Mia et al. [14]	CdSe (100 nm, RFS)/CdTe (CSS)	690	26.9	64.8	12.1
	CdS (140 nm, RFS)/CdTe (CSS)	847	24.7	70	14.6
	difference with CdSe	−157	+2.2	−5.2	−2.5
Paudel and Yan [20]	CdSe (100 nm RFS)/CdTe (CSS)	771	27.5	69.4	14.7
	CdS (15 nm, RFS)/CdSe (100 nm, RFS)/CdTe (CSS)	806	27.2	64.1	14.1
	CdS (130 nm, RFS)/CdTe (CSS)	811	24.2	75.5	14.8
	difference only CdSe	−40	+3.3	−6.1	−0.1
	difference mixed CdS/CdSe	−5	+3	−11.4	−0.7
pLingg et al. [8]	CdS (30 nm)/CdSe (60 nm)/CdTe (all HVE, substrate config.)	710	25.6	67	12.2
	CdS (120 nm)/CdTe (all HVE, substrate config.)	830	18.5	69.4	10.5
	difference with CdSe	−120	+7.1	−2.4	+1.7
Munshi et al. [15]	MgZnO (RFS)/CdTe0.8Se0.2 (800 nm, CSS)/CdTe (CSS)	854	28.4	79.1	19.1
	MgZnO (RFS)/CdTe (CSS)	860	26.3	78.9	17.9
	difference with CdSe	−6	+2.1	+0.2	+1.2
First Solar [3,9,10]	CdTe1−xSex/CdTe *	887.2	31.69	78.5	22.1
	CdS/CdTe *	875.9	30.25	79.4	21
	difference with CdSe	+11.3	+1.44	−0.9	+1.1

* The exact device structures and methods of deposition used in the First Solar record devices are not public knowledge.

, where the parameters are compared with reference devices without CdSe. Generally, the J_SC could be improved by the reduced band gap and by reducing or removing the CdS layer. V_OC losses are partially due to the reduced band gap. Further V_OC losses together with fill factor losses have been mostly attributed to interface recombination, but they were not further investigated. The highest V_OC was achieved by use of a MgZnO window layer instead of CdS. The deposition methods are also listed as they may have an influence on the material properties and device performance. The different approaches and results are discussed in the following subsections.

3.3. CdSe/CdTe Devices without a Window Layer

The performance of devices where the CdS window layer is completely substituted by a CdSe layer (Figure 2c) depends strongly on the thickness of the CdSe layer. Up to 100 nm thick CdSe layers completely interdiffuse with the CdTe during CdCl₂ treatment, forming photoactive zinc blende CdTe_1−xSe_x. This results in increased J_SC compared to CdS/CdTe devices (see Figure 3) on the one hand because of increased absorption in the long-wavelength region due to a reduced band gap of the absorber, and on the other hand because the absorbing CdS layer is eliminated (see Figure 4). The overall efficiency however is still reduced because of losses in V_OC and FF.

With thicker CdSe layers, the interdiffusion is not complete, and the J_SC is reduced due to parasitic absorption in residual non-photoactive wurzite CdTe_1−xSe_x with high Se content (x > 0.65). With optimized CdSe thickness, the J_SC can be increased by more than 3 mA/cm², however, this is not sufficient to offset the losses in V_OC, which are attributed to high interface recombination at the direct contact between CdTe_1−xSe_x and TCO [16,20].

The TCO used in CdTe and CdTe_1−xSe_x/CdTe devices is usually a bilayer of a low-resistivity material, such as SnO₂:F (FTO), and a high-resistivity material, such as SnO₂. The highly resistive transparent (HRT) layer is used to prevent direct contact between the absorber and the FTO in the case of incomplete CdS coverage [55]. Baines et al. [12] investigated different oxides as HRT layers for CdTe_1−xSe_x absorbers and found SnO₂ to be the best. It reduces V_OC losses that occur from direct contact of CdTe_1−xSe_x with FTO, but the interface recombination is still higher than with a CdS layer [12,20]. In fact, the devices listed in Table 3 all use a SnO₂ HRT layer.

With electron-beam induced current (EBIC) measurements the position of the junction can be determined: for up to 100 nm CdSe, the junction is located at the CdTe_1−xSe_x/TCO interface. This reveals that the CdTe_1−xSe_x does not act as a window layer, but as part of the p-side of the junction. For thicker CdSe layers, the junction is shifted into the CdTe_1−xSe_x layer, to the interface between photoactive and non-photoactive CdTe_1−xSe_x [16]. The V_OC and J_SC decrease for thicker layers, confirming that the presence of the non-photoactive wurzite CdTe_1−xSe_x is not beneficial for the device performance. Deposition methods and parameters therefore have to be chosen to minimize the formation of non-photoactive CdTe_1−xSe_x.

3.4. CdSe/CdTe Devices with a CdS Window Layer

With a combined CdS/CdSe window layer (Figure 2b), device performance can be increased by combining the upsides of both layers while minimizing the drawbacks (see Table 3). By reducing the CdS thickness from 130 nm in a CdS/CdTe device to 15 nm in a CdS/CdSe/CdTe device, the parasitic absorption in the short-wavelength region is minimized (region 1 in Figure 4a) and the J_SC is increased. Further J_SC increase is due to a shift in band gap with CdTe_1−xSe_x (region 2 in Figure 4a). The J_SC is therefore very similar to a CdSe/CdTe device, while the V_OC is improved with the CdS layer in between the CdTe_1−xSe_x and the TCO, which reduces the interface recombination (see Figure 4b) [20,54].

A CdS/CdSe/CdTe structure also works in substrate configuration solar cells (Figure 2e). Substrate configuration CdTe solar cells require a thick CdS layer in order to mitigate the formation of pinholes. With a CdTe_1−xSe_x layer between CdTe and CdS, the CdS layer thickness can be reduced from 120 to 30 nm with only small losses in V_OC and with a 7.1 mA/cm² increase in J_SC due to the reduction in CdS thickness and the reduction in band gap, resulting in an overall improvement in efficiency (see Table 3) [8].

3.5. CdTe_1-xSe_x/CdTe Devices with a MgZnO Window Layer

Recent high-efficiency CdTe devices have been achieved by replacing the CdS window layer with MgZnO (see Table 3) [15,22]. MgZnO has a band gap of 3.7 eV, higher than CdS, thereby reducing parasitic absorption, and it has a better band alignment with CdTe than TCO/CdS, reducing interface recombination and allowing for a higher V_OC [21,22]. However, introduction of a MgZnO window layer into a device structure can be problematic. The MgZnO/CdTe interface is sensitive to barriers originating from oxides, e.g., MgO, that can be formed during deposition. To obtain high-efficiency MgZnO/CdTe devices, deposition parameters, such as temperature and oxygen partial pressure during deposition, and CdCl₂ treatment, have to be carefully optimized [56].

Further improvement was achieved by using a CdTe_1−xSe_x layer at the front of the CdTe absorber. With an 800 nm thick CdTe_1−xSe_x layer in front of a CdTe absorber, Munshi et al. showed an improvement in device efficiency from 17.9% for a MgZnO/CdTe device to 19.1% in a MgZnO/CdTe_1−xSe_x/CdTe device (see Table 3) [15]. They improved the J_SC by 2.1 mA/cm² while retaining the high V_OC (see Figure 5a). As the device contains no CdS, the EQE in the short-wavelength region is improved (region 1 in Figure 5a).

The benefit of depositing CdTe_1−xSe_x instead of CdSe in front of CdTe is the low possibility that non-photoactive wurzite CdTe_1−xSe_x is formed. As a consequence, higher amounts of Se can be introduced to the device for better fine-tuning of the band gap gradient [29]. This results in a thicker photoactive low-band gap CdTe_1−xSe_x layer in the finished device. The low-band gap region appears sufficiently thick to absorb most photons with energies below the band gap of CdTe, down to about 1.42 eV, as evidenced by the shift in absorption edge in the EQE curve in region 2 in Figure 5a. The difference in absorption between the thick photoactive layer achieved with CdTe_1−xSe_x deposition and the thin photoactive layer achieved with CdSe deposition can be seen in the steepness of the EQE curve in region 2 in Figure 5a. The conclusion can be drawn that for maximized J_SC gain from CdTe_1−xSe_x, deposition of a CdTe_1−xSe_x layer instead of a CdSe layer is necessary.

The high V_OC in the MgZnO/CdTe_1−xSe_x/CdTe device was in part attributed to a high lifetime of 22 ns measured by TRPL, four times higher than in the MgZnO/CdTe device, due to improved interface quality between MgZnO and CdTe_1−xSe_x, and also due to the passivation of a recombination center by Se [13,15,19]. The MgZnO/CdTe_1−xSe_x/CdTe device has a much higher V_OC than the CdS/CdTe_1−xSe_x/CdTe devices. This indicates that the V_OC in these structures is limited by the TCO/CdS/CdTe_1−xSe_x interfaces, but further investigations into the V_OC and FF loss of CdTe_1−xSe_x/CdTe devices is needed to confirm this.

Figure 6 compares the EQE curve of the high-efficiency MgZnO/CdTe_1−xSe_x/CdTe device with the record device from First Solar. The First Solar device uses a CdTe_1−xSe_x layer and some kind of buffer layer instead of CdS [11]. It has a higher V_OC and higher J_SC than the MgZnO/CdTe_1−xSe_x/CdTe device (see Table 3) [3,11]. The increase in J_SC is primarily due to a very good absorption at high wavelengths (region 2 in Figure 6). The very straight absorption edge indicates the presence of a sufficiently thick CdTe_1−xSe_x layer with a band gap of 1.4 eV for complete absorption of photons in this region. Some further improvement of J_SC is achieved by an overall higher EQE, probably due to an antireflection coating. In the short-wavelength region (region 1 in Figure 6), the curves are very similar. This is an indication that First Solar must have used MgZnO or a similar material as a buffer layer, eliminating the influence of parasitic absorption. The increase in V_OC might be due to better interface engineering and therefore reduced recombination, but without more information on the First Solar device structure no conclusions can be drawn.

4. Conclusions

CdTe_1−xSe_x has been successfully introduced into CdTe devices by several different groups. It has been shown that a band gap gradient can be achieved by several different techniques and deposition methods. For better control and a higher limit to the amount of Se put into the devices, use of a CdTe_1−xSe_x layer is preferable to a CdSe layer in order to avoid formation of non-photoactive wurzite CdTeSe. In devices with a CdTe_1−xSe_x layer at the front of the CdTe absorber, the reduced band gap and possibility of reduced CdS window layer thickness result in increased short-circuit current. The open circuit voltage can be maintained at the level of CdTe devices due to Se passivating a recombination center in the absorber. The best reported devices use a MgZnO window layer instead of CdS, which provides a better band alignment with the absorber and eliminates parasitic absorption from the window layer.