Photovoltaic cell

The present invention relates to a photovoltaic junction having multiple quantum wells in the intrinsic region and a photovoltaic cell including such a junction that generates electricity from incident light such as sunlight. In particular, but not limited to, the junction is extended to longer wavelengths by using multiple quantum wells / barriers that are configured to receive highly focused sunlight and maintain stress equilibrium in the intrinsic region. It can be used as an upper junction in a tandem multi-junction solar cell with a GaAs-based lower photoelectrostatic junction with an absorbed end.

US2005 / 0247339 discloses the tandem solar cell shown in FIG. 7 of the document, which is shown here in FIG. The solar cell shown is a GaAs substrate 10 (or active or passive Ge substrate), a lower photovoltaic junction 12, which is formed on the substrate and configured to generate electricity from photons with wavelengths up to about 1040 nm. And includes an upper photovoltaic junction 16 formed on top of the lower junction and configured to generate electricity from photons with wavelengths below about 720 nm. Since the two photovoltaic junctions are connected by a tunnel junction 14, the same photocurrent passes through both junctions and is coupled into the electrical circuit by electrodes located at the bottom of the substrate and above the upper junction.

The lattice constant of each semiconductor layer is set to the lattice constant of the lower layer, namely GaAs or Ge, so that multiple semiconductor layers can be layered on top of each other with minimal defects that can significantly reduce the efficiency of the device. It must match the lattice constant of the substrate. It is well known in many areas of semiconductor technology that this strongly constrains the materials that can be used in each layer, but some deviations from this constraint are possible as described below. ..

The lower junction 12 can generate electricity from photons over most of the solar spectrum at wavelengths up to about 1040 nm with high quantum efficiency. However, much of the power from the shorter wavelength photons is during the junction due to the rapid heating of the free charge carriers, as the longer wavelength absorption ends generally produce photocurrents that occur at lower junction voltages. So, it is lost to the band gap energy corresponding to the absorption edge. In order to capture shorter wavelength photons at a higher voltage, the upper junction is composed of shorter wavelength absorption edges. This main structure is behind the development of tandem multi-junction solar cells with three, four, or more junctions, and is trying to capture the photocurrent of each junction at the maximum voltage. The limitation in such a multi-junction tandem cell is that when the junctions are laminated so that the long-wavelength light penetrates to the junction near the appropriate band edge where the longer-wavelength light is absorbed, the photocurrent of the entire stack is Must be the same as the photocurrent passing through each junction. The device is therefore limited by the junction that produces the minimum photocurrent.

Using the standard solar spectrum and the characteristics of each junction, it is possible to calculate the expected efficiency of a multijunction tandem device that depends on factors such as the absorption edge of each junction. An example contour diagram of efficiency for the lower junction bandgap and the upper junction bandgap (which is closely related to the absorption edge), taken from US2005 / 0247339, and the tandem cell in Figure 1. It is related to the same battery as, but here it is reprinted in Fig. 2. The characteristics of the lower cell formed from the GaAs pn junction and the upper cell formed from the GaInP pn junction (GaInP is a composition lattice-matched to GaAs) are shown by a thick crosshair. It can be seen that for optimum efficiency, the bandgap of both junctions must be as low as the absorption edge at longer wavelengths. The strong diagonal structure of the contour diagram is caused by the need to match the photocurrents produced by the two junctions.

In the sequence shown in FIG. 1, absorption of the lower junction by a single GaAs / GaAs pn junction by using an intrinsic region 18 containing a thin layer (quantum well) of InGaAs material with an absorption edge of about 1040 nm. The edges extend to wavelengths longer than about 880 nm, but quantum effects in thin wells change this value slightly. The InGaAs material is not lattice matched with the GaAs substrate, but each quantum well is located between the barrier layers of the GaAsP material that compensate for the lattice mismatch. If the quantum well and barrier are thin enough, and if the lattice constant, elastic modulus, and thickness of the layer are balanced, the intrinsic region can be aligned with the lattice constant of the GaAs substrate at extremely low anomalous levels. .. A "stress equilibrium" technique for achieving such a condition is described in WO 03/012881.

The upper GaInP cell is also configured with an intrinsic region containing quantum wells with increased indium concentration and absorption edges shifted to longer wavelengths. Since the increased lattice constants in these quantum wells are balanced by the intervening barriers with reduced indium concentration, the intrinsic regions formed from the barriers and wells are the lattice constants of the adjacent GaInP bulk regions and the underlying layer. It has a lattice constant that matches the lattice constant of the GaAs substrate.

In order to maintain an effective built-in electric field in the intrinsic region, the concentration of doping impurities must be kept extremely low so that the charge carriers generated by light are separated by the electric field rather than by diffusion. Must be. Lower impurity concentrations can maintain a built-in electric field in the thicker intrinsic region, but the practical lower limit of impurity concentration limits the thickness of the intrinsic region to about 500 nm to 1500 nm. To. According to the method described in WO 03/012881, the well / barrier intrinsic region can be constructed to such a thickness without significant defects caused by internal stresses, but the absorption cross section of the quantum well is that. Thicker quantum wells and thinner barriers are desirable because they depend on the total thickness of the wells integrated in the intrinsic region, which has a limited thickness.

From this point of view, the use of quantum barrier / well intrinsic regions in the upper junction formed from GaInP poses a problem. In order to match the lattice constant of GaAs, the gallium / indium ratio in GaInP must be about 0.51 / 0.49, resulting in an absorption edge of about 670 nm. To properly extend the absorption edge in the quantum well layer to about 720 nm, the gallium / indium ratio needs to be about 0.4 / 0.6, at which point the barrier balances the stress at about twice the thickness of the well. Achieve. For many spectral conditions, wavelengths longer than 720 nm, which require a significantly wider barrier, are desirable. This runs out of valuable thickness in the intrinsic region and reduces the potential total thickness, i.e. the absorption cross section of the quantum well material. Furthermore, stress equilibrium conditions need to limit the choice of quantum well thickness and barrier thickness that can achieve a constant absorption edge. To design a quantum well system, it is convenient to have more flexibility in the options.

In addition, of the group III gallium and indium materials, which tend to form a region on the adjacent (111) plane of gallium and indium atoms, using the continuous barrier and well layer of the GaInP-based material as a complete solution. It is difficult to form rather than mix well in the crystal structure without undesired ordering.

Therefore, it is desirable to provide another barrier / well structure based on GaInP with an extended wavelength absorption edge, and more particularly such a structure that can be formed as part of a device on a GaAs or Ge substrate. With such a structure, it is also desirable that the quantum wells in the structure can have a relatively large proportion of thickness and that the thickness of the wells and barriers can be selected more flexibly. It is also desirable that such a structure allows for greater reliability, defect-free growth, and more reliable and messy mixing of Group III species.

<p><patcit num="1"><text>US2005 / 0247339</text></patcit><patcit num="2"><text>WO 03/012881</text></patcit></p>

The present invention provides a photovoltaic junction having an intrinsic region containing continuous quantum wells formed of InGaAsP quaternary material. Such compounds can extend the absorption edge obtained using the ternary GaInP material to longer wavelengths.

It has been clarified that this quaternary material can be formed by mixing well with group III atoms and can provide high material quality that is important for forming quantum well photovoltaic junctions for high-efficiency operation. .. By properly adjusting the proportions of complementary In and Ga and the proportions of complementary As and P, the absorption edge of the material can be adjusted as needed over a considerable wavelength range, for example, the tandem solar. When used in a battery, it is possible to provide a well material having a desired lattice constant while adjusting the bonding performance. In this way, for example, a material that is tightly lattice-matched to GaAs (or other substrate such as Ge) with a band end adjustable to wavelengths from about 700 nm to 800 nm, which will be described in detail later, can be easily obtained. The range of about 700 nm to 740 nm is particularly useful in some of the described applications. Similar materials can also be obtained that are not lattice matched to the substrate but have a compensating barrier structure and are used in structures that maintain stress equilibrium.

Preferably, at least some quantum wells have an As to P relative ratio of As._yAnd P_1-yHowever, it is produced from an InGaAsP quaternary material defined by 0.2 &lt;y &lt;0.5, contains an InGaAsP quaternary material, or contains a layer of one or more InGaAsP quaternary materials, more preferably 0.25 y . It is 0.45, and in each case, it is arbitrarily lattice-matched with GaAs or other substrate materials. In the prior art, it was predicted that such bulk materials would not mix, and experiments have shown (K.Onabe, Jpn.J.Appl.Phys., 21,797,1982 and experiments by the present inventors). .. However, quantum wells, as evidenced by high photoluminescence signals, narrow photoluminescence peak widths, low background impurity levels, and dark current characteristics characteristic of materials dominated by luminescence recombination at high irradiation intensities. It has now become clear that this material can be produced with good material quality in the structure. It is observed that the intrinsic region with these quantum wells up to and above 50 remains depleted.

The composition of the well material (as well as the barrier material) should typically be lattice aligned, for example, within 2% of the substrate lattice constant, but this depends on the degree of stress equilibrium and / or other methods used. .. For reference, a reasonably accurate lattice alignment that can construct dozens of quantum wells can be within about 0.1% of the substrate lattice size, more preferably about 0.03%. When stress equilibrium is used with dozens of quantum wells, the quantum well / barrier structure must have a calculated independent lattice constant subject to similar constraints.

The present invention can also be defined by requiring the monolithic growth of the photoelectrostatic junction discussed, which may or may not include a monolithic layer intervening between the junction and the underlying substrate. It may be, for example, Ge, GaAs, or other material with similar lattice constants, including virtual substrates, Si or Ge, insulators, and the like. According to one embodiment, a_sIs the lattice constant of the substrate surface with the device on top, a_GaAsIs the lattice constant of GaAs, and a_GeIf is the lattice constant of Ge, then the matching of GaAs or Ge, respectively,<maths num="1"><img file="JP2012504331A_D0001.tif" /></maths>Or<maths num="2"><img file="JP2012504331A_D0002.tif" /></maths>Will be. This is consistent with the 0.1% state above and can be adjusted to reflect the 0.03% and 2% values above as well.

For efficient operation, the thickness of the quantum well is preferably at least 2 nm. Similarly, its thickness is preferably less than about 15 nm, more preferably less than 10 nm. The number of quantum wells can be configured to provide the desired amount of light absorption and current. However, preferably at least 15, more preferably at least 20 quantum wells are used.

Quantum wells are separated by a barrier layer. Conveniently, these may be ternary GaInP materials. Alternating barrier layers and well layers can then be grown by adding or not adding arsenic sources using epitaxial techniques. Since arsenic atoms are usually much easier to incorporate into epitaxial growth than phosphorus atoms, much less arsenic feed material is needed than phosphorus feedstock, with only minor changes in the amount of arsenic feedstock. Well, therefore, the proportion of arsenic incorporated can be precisely and easily controlled. Barriers can more generally be formed from Ga, In and P-containing semiconductor materials, including, for example, a quaternary material further incorporating Al, As, N or Sb.

An alternative material for wells between GaInP barriers that can extend band-end wavelengths in lattice-matched junctions to GaAs or Ge is GaAs. However, to construct such continuous barriers and wells, it is necessary to add or not add a fairly large flow of phosphorus raw material. Moreover, the built-in voltage of the photovoltaic junction is therefore too low for many purposes because the GaAs well provides a much longer wavelength band end than is possible with InGaAsP. Moreover, the band ends of the well are not adjustable with this material.

The lattice constant of the well material may differ from the lattice constant of the barrier material, in which case the well and the barrier must maintain a stress balance with respect to the substrate, i.e. the mutual compressive stress between the barrier and the well. The tensile stress must be balanced at a normal grid spacing that matches the grid spacing of the underlying substrate (and / or the grid spacing of the bulk semiconductor regions above and below the well and barrier). In order to avoid anomalies and improve the material quality of many quantum wells, for example 30 or more, this stress equilibrium requires consideration of the elastic modulus of the material as well as the layer thickness and lattice constant.

However, conveniently, the well and barrier materials can be selected so that the lattice constants of the well material and the barrier material both substantially match the lattice constants of the substrate. This allows the thickness of the well layer and the barrier layer to be selected regardless of stress equilibrium constraints. This makes it possible to increase the total thickness of the quantum well in the intrinsic region.

Barriers can be formed from ternary GaInP material and wells can be formed from quaternary InGaAsP material, where both materials can be formed, especially if the underlying substrate is GaAs or Ge, or if the top surface is lattice matched with GaAs or Ge. Has a composition that produces barriers and wells that are substantially lattice matched with GaAs or Ge. Conveniently, bulk semiconductor layers, one or both above and below the intrinsic region, also from GaInP materials, more generally quaternary materials incorporating a fourth element such as Al, As, N or Sb, etc. It is a Ga, In and P-containing semiconductor material, and may be formed from a semiconductor material lattice-matched with GaAs or Ge, if necessary. Further, a virtual substrate may be used to provide a lower layer substrate having an appropriate lattice spacing.

According to one variant, the InGaAsP well may be formed as a stepped well by incorporating a GaAs subwell separated from the barrier layer (usually GaInP) by a quaternary InGaAsP material. .. By including the GaAs subwells to form stepped wells of InGaAsP-GaAs-InGaAsP, the band ends of the structure can be extended to wavelengths of about 850 nm. The InGaAsP layer helps stabilize the transition from As-free GaInP to P-free GaAs.

Preferably, the junction is monolithically formed with the solar cell substrate. The joint may be formed monolithically with one, more, or all other tandem joints of the device.

The present invention also provides a solar cell comprising the above-mentioned photovoltaic junction, a semiconductor substrate, and electrodes arranged to connect photocurrent and photovoltage from the junction to the load.

Solar cells may include two, three, four or more photovoltaic junctions, especially photovoltaic junctions arranged in a tandem configuration so that the same photocurrent passes through all junctions. it can. In such an arrangement, the ability to adjust the absorption band ends of one or more junctions is important to match the photocurrent of each junction under typical irradiation conditions, and one or more junctions. Utilization of the InGaAsP well inside contributes to this.

In particular, a photovoltaic junction with wells of quaternary InGaAsP material can provide an upper junction arranged to receive incident light from the upper layer, and a solar cell can provide light from the upper junction. Further includes at least one lower joint arranged to receive the. The light received by the lower junction may be incident light that was not absorbed by the upper junction, but the photons emitted by the upper junction by the process of luminescence recombination of the charge carriers in the quantum well. It may also be included.

Generally, the lower junction is characterized by an absorption edge that has a longer wavelength than the upper junction and is based on a GaAs material such as GaAs, GaAsP, InGaAs. In particular, the lower intrinsic region can include an intrinsic region with InGaAs quantum wells separated by a GaAsP quantum barrier. Suitable solar cells can use approximately 20 to 100 InGaAs quantum wells that maintain stress equilibrium with the GaAsP barrier and match the lattice constants of the underlying GaAs or Ge substrate. The lower junction well is characterized by an InGaAs material preferably having an absorption band edge wavelength greater than 1000 nm (comparable to the band edge of a normal GaAs junction of about 875 nm), more preferably an absorption band edge wavelength greater than 1020 nm. Be done. Forming such a lower junction is typically compared to a conventional GaAs optoelectric junction where the upper junction of a regular GaInP provides insufficient photocurrent to match. It is possible to provide a significantly increased photocurrent with specific irradiation. Therefore, matching photocurrents can be provided by using quantum wells of quaternary InGaAsP material during the upper junction, which can be further adjusted by adjusting the composition and well thickness, solar cells. Can be optimized for a wide range of irradiation conditions.

The above photovoltaic junctions and solar cells can be used particularly as condensing cells under high irradiation conditions where sunlight is focused, for example, at least about 50 times, more preferably at least 100 times, on the surface of the solar cell. Activate. To this end, the invention also includes one or more solar cells as described above, one or more solar concentrators arranged to direct focused sunlight to one or more solar cells. Provided is a device including an electric circuit arranged to collect electric power from a solar cell.

The present invention is also a method of forming a photovoltaic junction using epitaxial growth techniques, such as MOVPE, in a reaction chamber, in which a first bulk semiconductor layer is formed, with a barrier layer and a quaternary InGaAsP above the first bulk layer. Provided is a method of forming an intrinsic region containing alternating quantum well layers of a material and forming a second bulk semiconductor layer above the intrinsic region. Barriers, quantum wells and bulk semiconductor layers can have various combinations of the above properties and the junctions may be formed as part of a multi-junction photovoltaic device. In particular, the barrier layer may be formed from GaInP, or one or both bulk semiconductor regions may be formed from GaInP, and the substrate may have GaAs or Ge or anterior surface, or the lattice constant presented may be GaAs or Ge. It may be a virtual board or other board that substantially matches the constant.

Preferably this junction, and preferably some other photovoltaic junction, is also monolithically formed with the substrate to form a monolithically grown solar cell device, but of course other steps such as metallization are usually performed. Needed.

The present invention may also be applied to other than solar cells and other photovoltaic junctions, such as LEDs, lasers, photodiodes, thus providing electronic and / or optics with pin junctions, wherein this junction is provided. The true region of contains continuous quantum wells formed from InGaAsP quaternary materials. Preferably, the quantum well has other properties, such as appropriate and lattice matching with the GaAs substrate as described above.

An embodiment of the present invention will be described below with reference to the drawings as an example.

<figref num="1">It is a figure which shows the tandem type solar cell discussed in US2005 / 0247339.</figref><figref num="2">It is a figure which shows the efficiency of various band gaps of the top and bottom junctions of a tandem type solar cell similar to the solar cell of FIG. 1 under typical solar irradiation.</figref><figref num="3">It is a figure which shows the structure of the tandem junction type solar cell of 1st Embodiment of this invention.</figref><figref num="4">It is a figure which shows the structure of the tandem junction type solar cell of the 2nd more detailed embodiment of this invention.</figref><figref num="5">Based on GaInP, it is a diagram of the measured and predicted dark current of the upper optical junction of the second embodiment, which contains 22 quantum wells of InGaAsP in the intrinsic region, and is more than at a bias of about 1.4 V (almost the working voltage of the cell). Both the steep slope and the proximity to the prediction of the curve indicating the "radiative limit" show good material quality consistent with the dominance of the radiative combination.</figref><figref num="6">Diagram of measured and predicted dark currents of optical junctions containing five GaAs quantum wells in the intrinsic region, based on GaInP. Significantly higher dark currents (about 100 times at 1 V bias) show much higher recombination and lower cell efficiency than the cells in Figure 5.</figref><figref num="7">It is a figure of the measured and predicted internal quantum efficiency of the device of FIG.</figref><figref num="8">It is a figure of the measured and predicted internal quantum efficiency of the device of FIG.</figref><figref num="9">Diagrams of the internal quantum efficiencies of the equipment in Figures 5 and 7, with predicted values for the lower junction containing 50 GaAsP quantum wells, measured and predicted values for the combined lower and upper junctions, and conventional tandem equipment. The measured value of the upper cell is also shown.</figref><figref num="10">It is a figure which summarized the epitaxial growth system of the multiple quantum well structure which embodied this invention.</figref><figref num="11">It is a figure which shows the modification of the InGaAsP quantum well system which prepares the subwell of GaAs and forms the stepwise structure.</figref><figref num="12">It is a figure which shows the condenser system used for the said photocell.</figref>

The structure of the solar cell according to the first embodiment of the present invention is schematically shown with reference to FIG. Not all layers are shown in the figure. Roughly speaking, the cell contains a substrate 20 on which a lower photovoltaic junction 22 is formed. The tunnel junction 24 is formed on the lower junction 22 and the upper photovoltaic junction 26 is formed on the tunnel junction 24. The structure above the upper photovoltaic junction may include a window layer 28, with electrodes 30 usually located above the window layer and below the substrate. The solar cell operates by supplying irradiation 32, generally focused solar irradiation, to the upper part of the battery, and draws out the generated electric power by connecting to the electrode 30. The upper photovoltaic junction absorbs incident photons with higher energy than the band edge of the junction. Lower energy photons pass through the lower junction with the lower energy band end. Higher energy band edge materials usually provide higher junction voltage, so incident light photons are therefore used to generate current at higher voltage if possible, while lower energy photons are on the underside. It is not collected and discarded by joining.

The lower and upper photovoltaic junctions are shown as pin junctions, respectively. In such a junction, the intrinsic region i is completely depleted during operation and the charge carriers generated in the intrinsic region by irradiation 32 are carried by the electric field of the junction to the adjacent p- and n-doped bulk semiconductor regions. Is generally intended. Of course, in this example and other examples, the doping polarity may be reversed as needed.

The intrinsic region of the upper junction is formed from multiple quantum wells, each well containing InGaAsP material.

The choice of other materials for the solar cell depends on various factors and constraints, and specific detailed examples are shown in the second embodiment below. In general, the substrate may be made of GaAs or germanium, and the lower photovoltaic junction may be a GaAs-based junction. In particular, the lower junction may be a pin junction and may have a bulk semiconductor region formed of GaAs. The true region of the lower junction may itself contain quantum wells, for example to extend the band ends of the junction to longer wavelengths. These wells and surrounding barriers can be stress-balanced or strain-balanced by the methods discussed in the prior art and elsewhere in this document, eg extending the band ends of the intrinsic region to longer wavelengths. To be able to. This is important for the device to make better use of the available incident light spectrum. There may be other photovoltaic joints above or below between the lower and upper joints in the device, and / or the lower joint may be omitted.

The quantum wells of the upper junction shown are placed between the quantum barriers. These barriers may be ternary GaInP or GaInP based, and one or both of the bulk p-region and bulk n-region of the upper junction shall also consist of GaInP or GaInP based quadrants. However, other materials may be used, in which case the perimeter of the intrinsic region becomes a more complex structure. In general, bulk p-region and n-region and quantum well and barrier materials may be lattice matched with the lower photoelectrostatic junction and the underlying GaAs or Ge substrate, but elsewhere in this document. Wells and barriers with strain / stress equilibrium may be used as needed as described. By using an InGaAsP quantum well in this structure, the band end of the intrinsic region can be extended to a longer wavelength. This can be used to control the photocurrent of the upper junction to better match the photocurrent of the underlying junction, such as the GaAs-based junction described above.

It has been found that the proportion of arsenic between about 0.25 and 0.45 is suitable for InGaAsP, especially for materials that are lattice matched with GaAs within about 2%. The proportion of Group III elements can be adjusted to a suitable lattice size, eg, to match a particular substrate, or to maintain a balance of barrier compositions that are complementary to the substrate.

One or more reflective structures, such as specular or Bragg reflective structures, may be included in the solar cell, which reflects photons of incident light or photons generated by emission recombination during junctions. Return to the joint. For example, such a structure can be placed between the lower photovoltaic junction and the substrate.

The structure of the second embodiment of the present invention will be described with reference to FIG. This is a more detailed example of the structure of the first embodiment and is omitted to clarify some aspects such as electrodes. This second embodiment can be adapted and used in various ways as described for the first embodiment. The individual details of the second embodiment can be applied to the first embodiment in various ways apparent to the technician. For example, the details of the upper photovoltaic junction of the second embodiment can be used in a variety of different solar cell structures commonly discussed with respect to the first embodiment.

The solar cell of the second embodiment is formed on the GaAs substrate 50 using the metalorganic vapor phase growth method (MOVPE), but other epitaxial techniques may be used. The main layers shown are, in order from the substrate, the lower photovoltaic junction, the tunnel junction, and the upper photovoltaic junction. As shown in the figure, the lower photovoltaic junction comprises an n-doped GaAs bulk semiconductor layer 52, a first multiplex quantum well (MQW) structure 54, and a p-doped GaAs bulk semiconductor layer 56. Then, the AlGaAs window layer 58 is provided, followed by the tunnel junction 60. The upper photovoltaic junction comprises an n-doped GaInP bulk semiconductor layer 62, a second multiplex quantum well structure 64, and a p-doped GaInP layer.

The lower multiple quantum well structure comprises dozens of thin layers, preferably more than 30 layers, of quantum well material having a longer wavelength band end than the surrounding bulk GaAs material. The band edges of the well material preferably extend to wavelengths above 1000 nm, such as 1020 nm, 1040 nm or 1100 nm, allowing better use of the incident solar spectrum as described in the prior art. This can be achieved, for example, in the various ways described in US2005 / 0247339 and US2003 / 0089392 (both incorporated herein by reference). In the second embodiment described herein, the lower junction well is formed using an InGaAs ternary material with a larger lattice constant than the surrounding bulk material and / or substrate. To avoid the formation of lattice defects that can increase carrier recombination during bonding to the extent that it reduces efficiency more than necessary, the quantum barrier layer between the wells is from a material with a reasonably smaller lattice constant, in this example GaAsP. Since it is formed using a ternary material, the quantum well structure has an overall lattice constant that matches the lattice constant of the substrate, considering the layer thickness, lattice parameters and elastic modulus of the material.

The upper quantum well structure is similarly constructed, but multiple quantum wells formed from the quaternary InGaAsP material between the barrier layers of the ternary GaInP material are used. In the second embodiment, both the upper junction well and barrier are made of a material that is lattice matched with the surrounding GaInP bulk region and the GaAs substrate, maintaining strain or stress equilibrium as in the first quantum well structure. Although not, such a method can be used as needed.

More specifically, the layer of the solar cell of the second embodiment can be provided roughly upward from the substrate as follows. -GaAs substrate-The use of a (100) substrate cut at 10 degrees on the 111 plane can be used in place of the normal (100) cut at 3 degrees on a [100] substrate, thereby in the GaInP layer. Increases the irregularity of group III atoms in (more than 2 degrees, preferably more than 7 degrees from the &lt;111&gt; or &lt;110&gt; crystal plane, is also an advantage). -1 x 10 with silicon atom¹⁸cm^-3100 nm thick GaAs layer n-doped to the density of -2 x 10¹⁷cm^-32000nm GaAs, n-doped with silicon atoms -50 quantum wells, each 6.8 nm thick In_0.23Ga_0.77As is Ga with a thickness of 28.6 nm each_0.89AsP_0.11Separated by a quantum barrier, the well has a half-thickness barrier on the outside of the terminal well, and the well is characterized by a 1016 nm photoluminescence peak with a half-width of 57.7 meV. -Non-doped 10nm GaAs, -2 x 10¹⁸cm^-3400nm GaAs emitter, p-doped with carbon atoms of -43nm Al_0.8Ga_0.2As p window, -15 nm Al densely p-doped with silicon atoms_0.45Ga_0.55As and a densely n-doped 15 nm GaAs tunnel junction, -30 nm AlInP, n-doped minority carrier reflector with silicon atom, -Lattice matched to GaAs, 1x10¹⁷cm^-3570nm GaInP, n-doped with silicon atoms -22 quantum wells, each 4.4 nm thick GaInAsP (composition see below) separated by a 15.4 nm thick GaInP quantum barrier, all lattice-matched (or very closely matched) to GaAs, The well has an X-ray period of 19.8 nm and is characterized by a 716 nm photoluminescence peak in the center of the 4-inch wafer used. -60nm GaInP lattice matched to GaAs, 1x10¹⁸cm^-3Dope with zinc atom, -40nm GaInP lattice matched to GaAs, 3x10¹⁸cm^-3Dope with zinc atom, -30 nm AlInP lattice matched to GaAs, 5 × 10¹⁷cm^-3Dope with zinc atom, -1 x 10²⁰cm^-3150 nm GaAs p-doped with carbon atoms.

Then, suitable electrode contacts and a protective layer are added as desired.

The proportion of In and P in the GaInAsP quantum well is not specified in the list of layers above. The band ends of the well, or solar cell properties, can be adjusted by varying these proportions, especially the relative ratio of indium to phosphorus, and the proportions of Ga and As are adjusted to grid-match the well to GaAs. Can be left alone. According to "Semiconductors: Data Handbook", O. Madelung, Birkhauser, 2004, pp. 167-168, GaInAsP Ga_{(1 + y) /2.08}In_{(1 + y) /2.08}As_yP_1-yAll values of y by can be matched with GaAs.

Values in the entire range from y = 0 to y = 1 provide a bandgap range for wells in the second quantum well structure from about 1.42 eV to about 1.9 eV. One suitable value for y for the present invention is about y = 0.33, which is Ga._0.66In_0.34As_0.33P_0.67The composition of is applicable. Another suitable composition for lattice matching to GaAs is Ga_0.705In_0.295As_0.4P_0.6It gives absorption up to about 724 nm to a 4.4 nm thick quantum well. Suitable for GaInAsP compounds with a y value of at least about 0.25 to 0.45, and optionally a value of about 0.2 to about 0.5, lattice matched to GaAs or Ge within a few percent (eg 2%). In this range, the quaternary GaInAsP quantum well matches the photocurrent of the lower GaAs-based junction without increasing the photocurrent of the upper junction moderately and significantly reducing the voltage output of the upper junction. Sufficiently increase the photocurrent of the upper junction. At y = 0.33, the band edge of the well is about 70 nm above the band edge of the bulk GaInP photovoltaic junction at about 650 nm, some over 40 nm. As implied in the discussion of Figure 2, if further control is used to optimize the solar cell for specific spectral conditions, the material composition of the lower junction well can also be adjusted to have similar properties to the band ends. Can be controlled.

FIG. 5 shows the dark current (black circle) measured in the laboratory test of the upper photovoltaic junction of the tandem device embodying the present invention detailed above. Specifically, the test device uses an InGaAsP quantum well that is substantially lattice-matched to GaAs (the test device is formed with a very slight strain balance between the well and the barrier, and the lattice constant of the intrinsic region is used as the substrate. The arsenic ratio is about 0.28 and the indium ratio is about 0.66 (accurate lattice matching requires an indium ratio of about 0.64). This composition is in the "immiscible gap" reported in the K. Onabe literature above, but Figures 5 and 7 show that it is normally a thin quantum well structure with high luminescence and thus also emission absorption efficiency. It shows that it can be formed.

The dark current is a diode current that is generated when the photovoltaic junction of the junction is not irradiated but is forward biased in the expected direction of diode conduction using a voltage in abscissa. In order to measure the dark current from the upper junction of the tandem type equipment, a test apparatus containing only the upper junction is formed.

Lower dark currents generally indicate improved solar cell performance, as the overall performance of the junction being tested can be simulated by open-circuit photocurrents minus dark currents. The higher forward bias produces a larger dark current in the device and thus also replaces the action under higher irradiation levels.

The equipment to be tested is also "Efficiency limits of quantum well solar cells", JP Connolly et al., 19th European Photovoltaic Solar Energy Conference, Paris, Simulated using the "SOL" model described in Proceeding, pages 355-359, June 2004. The only free parameter, the carrier lifetime, is adjusted to fit the model well to the data (the "theoretical total amount" shown by the solid line), where the lifetime has a value of 90.0 ns. The value of Rs shown in the graph is the series internal resistance of the device, about 10³A / m²Affects dark currents above and decreases the dark current density towards the linear function of the device bias (note that Figure 5 is a linear logarithmic plot). The model can also visualize the theoretical limits of performance for comparison, based on various hypotheses. The line containing the triangle marks shows the much lower current expected from Shockley's equation for ideal diode action that balances carrier diffusion against drift in the intrinsic pn junction electric field. Lines with circles indicate the current expected at the emission limit, taking into account the inevitable photorecoupling of charge carriers, which is the reverse process of charge carrier photogeneration. Lines with a plus mark also take into account non-emission Shockley-Read-Hall carrier loss due to recombination via trap states associated with lattice defects and similar defects.

In Figure 5, the emission limit and the ideal Shockley curve are separated by about an order of magnitude. Importantly, both measured and predicted dark currents approach and intersect the emission limit curve with high bias and high current densities. This cell is in this case under a high irradiation level equivalent to about 100-200 times standard solar irradiation, as the higher bias during dark current measurements replaces the higher irradiation levels under the working conditions of the solar cell. It can be seen that the theoretical emission limit of efficiency is approached. This is achieved by reducing the carrier loss through the non-emission mechanism to a negligible level, thus demonstrating that very good material quality of quantum wells and barrier materials and extremely high efficiency solar cells can be provided.

For comparison with FIG. 5, FIG. 6 shows an equivalent dark current curve of the test equipment showing much lower performance. The device incorporates a GaInP-based upper photovoltaic junction lattice-matched to GaAs, and the intrinsic region is replenished with five quantum wells formed from GaAs. As is clear from the quantum efficiency in FIG. 8, this absorbs photons with energies well below the 1.85 eV bandgap of GaInP. This additional absorption range is expected to provide superior photocurrents in the intrinsic region compared to GaInP-only structures. However, the very critical nature of the well (ie the low energy bandgap) also reduces the voltage of the junction and significantly increases the dark current compared to the junction in Figure 5, adversely affecting the overall performance of the solar cell junction. Means to receive.

It should also be noted that the quantum efficiency in Fig. 8 (peak less than 0.7) is much lower than the quantum efficiency in Fig. 7 (peak of about 0.95). This indicates the disappearance of the electric field due to the charged impurities in the true region.

In detail, the device in Figure 6 -1.0 × 10¹⁸cm^-3500 nm thick n-buffer of GaAs n-doped with -2.0 × 10¹⁸cm^-3Si-doped AlInP 60 nm minority carrier reflector layer, -1.0 × 10¹⁷cm^-3Si-doped GaInP in 570 nm bulk n-region, -Five-fold repeating layer of barrier and well, barrier is 60 nm undoped GaInP, well layer is 6 nm undoped GaAs, -100 nm thick i-region buffer, undoped GaInP, -7.0 × 10¹⁸cm^-3Zn-doped 100 nm bulk p-region, -Ten¹⁸cm^-3Zn-doped AlInP 30 nm window layer, and -2 x 10²⁰cm^-3Includes a 150 nm GaAs cap, C-doped in.

In Figure 6, the adapted carrier life is as short as 23 ns, and it is clear that dark currents above about 0.6 V are dominated by Shockley-Read-Hall carrier loss at all biases, material quality in quantum wells. Has been shown to be inferior and inadequate solar cells. High DC resistance used (290 Ohms / m²) Nevertheless, neither the predicted nor the measured dark current approaches the emission limit indicated by the circle, and the measured dark current is several orders of magnitude larger than the current according to the ideal Shockley curve even at high bias.

Figures 7 and 8 show the internal quantum efficiencies (black circles) measured in laboratory tests of the same equipment used in the dark current plots of Figures 5 and 6, respectively. Quantum efficiency is the probability that a photon of a predetermined wavelength will be absorbed in the device and the charge carriers generated will be recovered at the electrical contacts, resulting in the current passing through the cell. Therefore, the area under the quantum efficiency curve is proportional to the total photocurrent generated by the device when weighted by the wavelength dependence of the incident light (eg, the solar spectrum). Internal quantum efficiency ignores photon loss due to reflection on the top surface of the device. To measure the quantum efficiency of only the upper cell, use light of a wavelength that is absorbed only by the lower junction and of sufficient intensity that the lower junction photocurrent does not limit the device. Irradiate the device. That is, the measured quantum efficiency represents the performance of only the upper junction.

In Figures 7 and 8, simulations were performed using the SOL model described above. Simulations were used to calculate the contributions to quantum efficiency from each of the bulk p-region, bulk n-region, and intrinsic i-region. In both test devices, as expected, the contribution from the intrinsic region, where the electric field quickly and separately wipes out the newly light-generated charge carriers, is predominant. The plunge from the main peak in quantum efficiency occurs at the band edge of the bulk semiconductor material, which is GaInP, in each case, with lower steps continuing to higher wavelengths. This step is due to light absorption in quantum wells of photons that are too long to be absorbed in the bulk region, and the extension to longer wavelengths is much larger in GaAs wells than in InGaAsP wells (GaAs bandgap). As mentioned above, this can produce higher photocurrents (more photons are absorbed) than the lower junction voltage (the effective bandgap of the junction is smaller). The large number of InGaAsP quantum wells in 22 wells at 4.4 nm each in Figure 7 are relatively higher due to the larger total absorption cross section of the quantum wells compared to the five GaAs wells at 6 nm each in Figure 8. Bring a step.

A comparison device with five GaAs quantum wells is characterized by a relatively poor quantum efficiency of up to about 68%. This indicates the loss of the electric field due to impurities charged in the true region due to interface recombination, dislocations and other defects. On the other hand, the device with 22 wells of InGaAsP shows a peak quantum efficiency of over 95% dominated by the contribution of the i-region, again showing the excellent material quality in the quantum wells and barriers. As already mentioned, the size of the long wavelength steps in the quantum efficiency curve, i.e., can be easily adjusted by easily adjusting the elemental structure and composition of the quantum well region in the InGaAsP device to match the photocurrent of the lower junction. The total photocurrent of the upper junction can be controlled. In addition, this can be done in a wide selection of solar radiation, including predicted variations in both spectrum and intensity.

FIG. 9 shows the internal quantum efficiency measurement of the complete tandem cell detailed above, and the test results of only the upper junction for that purpose are shown in FIGS. 5 and 7. The quantum efficiencies (thin solid lines) calculated using the SOL model are shown separately for each of the lower and upper junctions, and also for the two coupled junctions. The measured internal quantum efficiencies (circles) of the upper cell and the two bonded junctions are shown. For further comparison, the internal quantum efficiency of the upper cells of known JEC tandem cells is shown by a thick solid line (Over 30% efficient InGaP / GaAs tandem solar cells). , T. Takamoto et al., Appl. Phys. Lett., 70, 381, 1997).

The step at wavelengths extending beyond the GaAs band end of about 880 nm to about 1040 nm is due to absorption in the 50 InGaAs wells of the lower junction. The increased photocurrent associated with this step coincides in the upper photojunction by a relatively smaller step beyond the GaInP band end of the quantum efficiency curve of the upper junction, and thereby also in its wavelength range. The decrease in photon utilization efficiency results in a corresponding downward trend in the quantum efficiency of the lower junction.

The multiple quantum well structure of the upper junction of the quaternary GaInAsP is particularly convenient for formation using epitaxial techniques such as MOVPE because the introduction of a controlled amount of arsenic source allows arsenic to be mixed into the material. This is because it can be controlled precisely. Phosphorus, which is another group V element, is difficult to mix in GaAs type materials, so a high concentration of phosphorus raw material or raw material (usually phosphine) is required. Arsenic is more easily mixed, so keeping the phosphorus source supply high and adding or not adding a finely controlled small amount of arsenic (usually arsine) alternates layers of GaInP and GaInAsP. Can be formed with the correct proportion of arsenic. In the detailed examples shown below, to form alternating layers of GaInP and GaInAsP, the phosphine flow into the MOVPE chamber is kept at a saturated phosphorus admixture level while arsine is switched at a much lower level. It is clear that this can provide wells and barriers, respectively. Higher supply concentrations and higher trimethylgallium, for example by opening and closing two complementary inlets, to maintain either lattice matching with GaAs or strain-balanced quantum well structures as needed. Switching between low feed concentrations forms wells and barriers, respectively, while keeping the trimethylindium feed constant at saturation levels.

A detailed overview of suitable MOVPE growth forming a structure similar to the second photovoltaic junction outlined above is given below. For the sake of clarity, the junction structure here will be described as being formed directly on a GaAs substrate. Of course, instead, it can be formed on top of tunnel junctions and / or other structures as needed, and MQW structures with both wells and barriers lattice-matched with GaAs or strain balanced. It is clear that you can choose whether to provide it. Further, if necessary, another structure may be formed on the upper part of the quantum well. The outline is shown in Fig. 10.

1) A (100) n-GaAs substrate deviated by 10 degrees from (111) A was first placed in a MOVPE reactor (Thomas Swan, flip top 7x2 showerhead) with a pressure of 100 Torr and a pure hydrogen carrier flow rate of 20 L. Operate at / min. The substrate is heated for 980 seconds in the presence of arsine at a flow rate of 150 standard cc / min (sccm) to a maximum pyrometer temperature of 720 ° C to 730 ° C (from 950 nm emission). During this time, the surface oxide is removed from the substrate.

2) Then cool the substrate to a pyrometer temperature of 670 ° C while maintaining the arsine flow rate at 150 sccm. 1.86 × 10^-2A molar fraction of trimethylgallium vapor (TMG) flow is introduced to form a 500 nm GaAs (buffer layer) at a vapor deposition rate of 0.83 nm / sec.

3) Then replace the arsine flow with 250 sccm of phosphine and at the same time the buffer layer TMG flow 4.2 × 10^-3Replaced with another TMG source of mole fraction, 3.9 x 10^-3A 15.4 nm GaInP barrier layer is deposited with a molar fraction of trimethylindium flow to form tensile strained GaInP (53.3% Ga, 46.7% In) at a formation rate of 0.38 nm / sec. Alternatively, GaInP can be lattice matched to the composition of 51.3% Ga, 48.75% In using an appropriate TMG flow.

4) Compressively strained quaternary QW was simultaneously added to the reactor with 2.3 sccm of arsine using a 2-fold diluted arsine stream in hydrogen, and TMG was 7.7 × 10^-4It is formed by exchanging for another TMG source of mole fraction, resulting in a gallium ratio of 64% to the quaternary alloy. Compressive strained x = 0.64, y = 0.32 quaternary alloy (mismatch 2.1 × 10)^-3(Compression) is deposited for 10.3 seconds to form a 4.4 nm thick quantum well layer. When the barrier is lattice-matched GaInP, the conditions of the quaternary composition are adjusted to obtain a lattice-matched composition.

5) Then, the next barrier layer of GaInP is deposited by returning the gallium fraction to 53.3%, removing arsine from the reactor, and exchanging the TMG source. The MQW to be generated is 5x10^-4It has a residual tensile strain of, which can be optimized for a true strain equilibrium structure by fine-tuning the growth temperature or quaternary TMG flow.

6) Steps Steps 4 and 5 are repeated to create a strain-corrected, normally 37 GaInAsP / GaInP period multiple quantum well structure. Alternatively, GaInP and GaInAsP can be lattice-matched.

FIG. 11 shows another modification of the structure of FIGS. 3 and 4 above. According to this variant, by forming GaAs subwells 70 in some or all quaternary InGaAsP quantum wells 72, layers of InGaAsP each GaAs subwell 70, GaInP three as shown in the figure. Separate from each barrier layer 74, which may be made of the original material. As described above, the quantum well structure may be lattice-matched, or the portion such as the InGaAsP layer may contain a slight strain.

The quaternary composition is an alloy that creates a quantum well with an absorption band end at a wavelength of about 735 nm using a GaAs substrate, as described above, or in certain examples, ie Ga._0.62In_0.38As_0.34P_0.66Close to. The thickness of the quaternary layer must be small enough so that it does not split. The InGaAsP layer provides a GaInP to GaAs transition material, which contains As, thus improving the GaInP to GaAs interface quality. Quantum well structures that transition directly between GaInP barriers and GaAs wells have boundary changes from As and non-P to P and non-As, probably relatively low due to the non-luminescent recombination center at the well-barrier boundary. Indicates the luminescence rate.

The typical size of the stepwise multiple quantum well structure in FIG. 11 can be 1.7 nm in the InGaAsP layer and 2.0 nm in the GaAs layer, and emits light at about 850 nm. The quaternary composition as the only barrier material and the GaAs-based multiple quantum well structure as the well material do not provide confined electron states. Therefore, the GaInP barrier layer is required to activate the sequence.

The table below details the structure of the device or junction that can form the stepped well arrangement of FIG. The table shows layer thickness, doping type, cm^-3Indicates the doping density represented by, and if appropriate, the photoluminescence wavelength of the layer represented by nm.

<tables num="1"><img file="JP2012504331A_D0003.tif" /></tables>

The above photovoltaic junctions and solar cells can be used especially for concentrating solar cell applications, where sunlight is usually focused about 50 to 1000 times, usually only about 0.1 mm.²From about 200mm²Guided to a condensing solar cell with a surface area of. Such a system, shown in FIG. 12, may include, for example, a condensing element 80, a plurality of solar cells 82, and a power management circuit 84 for collecting and appropriately forwarding power from the solar cells. The tracking system 86 may be used to properly point the concentrator toward the sun so that the concentrating solar cell operates with the optimum efficiency possible. A cooling device 88 may be required to maintain the solar cell at an appropriate operating temperature. In FIG. 12, a computer-based control unit 90 centralizes the management of various subsystems.

Although detailed embodiments have been shown, it is clear that various modifications can be made without departing from the scope of the invention. For example, quantum wells, barriers and other layers of various materials and thicknesses have been mentioned, but this is one or more wells with different compositions and / or thicknesses, within the range of the structures described. Or it does not preclude the addition of layers of different properties such as barriers. Where binary and ternary semiconductor materials are mentioned, modifications of these materials may be used, including suitable ternary and quaternary materials, if appropriate.

Bulk semiconductor regions formed with GaInP are also closely related quaternary compounds, including, for example, GaInP-based quaternary materials with similar lattice size and / or bandgap properties, separately containing Al, As, N or Sb. It may be formed from other materials such as, in which case it becomes the upper cell barrier structure.

Although a part of the substrate such as GaAs has been described, other substrate materials or a method such as a virtual substrate or a relaxed substrate may be used if necessary. Similarly, bulk semiconductor layers of different materials, properties and thicknesses may be used, including inversion of p- and n-doping schemes.