Microelectrical characterizations of junctions in solar cell devices by scanning Kelvin probe force microscopy

ARTICLE IN PRESS Ultramicroscopy 109 (2009) 952–957

Contents lists available at ScienceDirect

Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic

Microelectrical characterizations of junctions in solar cell devices by scanning Kelvin probe force microscopy C.-S. Jiang a,, A. Ptak a, B. Yan b, H.R. Moutinho a, J.V. Li a, M.M. Al-Jassim a a b

National Renewable Energy Laboratory, 1617 Cole Blvd., Golden, CO 80401, USA United Solar Ovonic LLC, 1100 West Maple Road, Troy, MI 48084, USA

a r t i c l e in f o

PACS: 72.40.+w 68.37.Tj 73.30.+y Keywords: Atomic force microscopy Microscopic methods Specifically for solid interfaces and multilayers

a b s t r a c t Scanning Kelvin probe force microscopy was applied to the microelectrical characterizations of junctions in solar cell devices. Surface Fermi-level pinning effects on the surface potential measurement were avoided by applying a bias voltage (Vb) to the device and taking the Vb-induced potential and electric field changes. Two characterizations are presented: the first is a direct measurement of Bi-induced junction shift in GaInNAs(Bi) cells; the second is a junction-uniformity measurement in aSi:H devices. In the first characterization, using Bi as a surfactant during the molecular beam epitaxy growth of GaInNAs(Bi) makes the epitaxial layer smoother. However, the electrical potential measurement exhibits a clear Bi-induced junction shift to the back side of the absorber layer, which results in significant device degradation. In the second characterization, the potential measurement reveals highly non-uniform electric field distributions across the n–i–p junction of a-Si:H devices; the electric field concentrates much more at both n/i and i/p interfaces than in the middle of the i-layer. This non-uniform electric field is due possibly to high defect concentrations at the interfaces. The potential measurements further showed a significant improvement in the electric field uniformity by depositing buffer layers at the interfaces, and this indeed improved the device performance. & 2009 Elsevier B.V. All rights reserved.

1. Introduction Junction formation is critical in solar cell devices, because the junction is one of the core structures in determining device performance [1]. The electrical properties of the junctions (such as location, depth, and electrical potential distribution) are measured and controlled in high-performance solar cell design and fabrication. Secondary ion mass spectrometry (SIMS) that measures donor and acceptor concentrations is a conventional method often used to determine the junction depth [2,3]. However, although SIMS measures dopant concentrations, it does not directly measure carrier concentrations. If carriers originate not only from foreign dopant specimens but also from intrinsic defects, SIMS cannot provide an unambiguous measurement of junction depth. High-performance solar cells often possess absorbers with very high purity or low carrier concentrations to extend the depletion and carrier collections. The carrier concentrations are often lowered to 1013–1014 cm3, which is close to, or below, the SIMS measurement limits for dopant elements. On the other hand, conventional electrical measurements such as the capacitance method have been widely used to characterize

 Corresponding author. Tel.: +1 303 384 6687; fax: +1 303 384 7690.

E-mail address: [email protected] (C.-S. Jiang). 0304-3991/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2009.03.048

junctions [4,5]. The carrier concentration and depletion width can be determined by capacitance–voltage (C–V) measurements. However, this method does not provide information about the junction location or depth, both of which are key factors in determining device performance. Scanning Kelvin probe force microscopy (SKPFM) [6–8] measures two-dimensional surface potential, with resolutions in the order of nanometers, and provides direct measurements of the electrical potential on junctions of solar cell devices [9–11]. In this paper, we present two examples of such junction characterizations: the first is a direct observation of Bi-induced junction shift in GaInNAs(Bi) solar cells; the second is an investigation of the electric field uniformity in hydrogenated amorphous silicon (a-Si:H) cells. These characterization studies are helpful toward understanding the device physics and potential routes to improving solar cell designs. 2. Experiments SKPFM is based on the non-contact mode of atomic force microscopy (AFM), which uses atomic force between the AFM tip and sample surface for the topography measurement. In addition to the atomic force, there is also a Coulomb force if the contact potential between the tip and the sample is different, which happens either in case of different work functions between the tip

ARTICLE IN PRESS C.-S. Jiang et al. / Ultramicroscopy 109 (2009) 952–957

3. Bi-induced junction shift in GaInNAs(Bi) solar cells GaInNAs with a few percent of N and In is considered promising for a lattice-matched four-junction solar cell, as the third cell [12], but roughness caused by In segregation leads to poor epitaxial qualities [13,14]. To circumvent this problem, a small amount of Bi surfactant was used during the MBE growth, which effectively improved the epitaxial layer smoothness, as seen previously [12–14]. However, the Bi incorporation induced serious device degradation, especially in the short-wavelength range of quantum-efficiency measurements, regardless of its improvement on the epitaxial quality. A quantum-efficiency simulation indicates a possible junction shift or a type reversion from p-type to n-type in the absorber with the addition of Bi [12]. Here, we report on a direct measurement of the junction locations using SKPFM. We will first show the effect of Bi as a surfactant on the epitaxial quality and then focus on the Bi-induced junction shift. Fig. 1 shows the surface root-mean-square (RMS) roughness

2.5

2 RMS Roughness (nm)

and sample or when a bias voltage is applied between them. This Coulomb force is used for the surface potential measurement in SKPFM. The SKPFM technique was first developed by Nonnenmacher et al. [6] in the early 1990s, and was significantly improved by Hirahara et al. by electrically stimulating an oscillation mode at the second resonant frequency of the cantilever [7]. Detailed descriptions of the technique can be found in the literature [6–9]. The GaInNAs(Bi) solar cell devices were grown by molecular beam epitaxy (MBE) with elemental sources of Ga, In, and Bi; an arsenic cracker; and a radio frequency (RF) plasma source for atomic nitrogen, as described in detail in Ref. [12]. The GaInNAs(Bi) samples subjected to the SKPFM measurement were cleaved along the [11 0] direction, and the junctions in the devices were exposed onto the (11 0) cross-sectional surface. The cleaved surfaces are atomically flat with a few steps in heights of several atomic layers. Most III–V devices can be cleaved atomically flat, and thus are suitable for SKPFM measurement. The devices were connected to a voltage source, and a bias voltage (Vb) was applied between the front and back contacts of the devices during the SKPFM measurement. Positive and negative Vbs represent forward and reverse biases, respectively. The a-Si:H devices were fabricated using RF glow discharge on Cr-coated GaAs(0 0 1) wafer substrates, in a multi-chamber system for high-performance n–i–p solar cell fabrications. The n-layer was a phosphorus-doped a-Si:H with a thickness of 20–30 nm, and the p-layer was boron-doped nanocrystalline silicon (nc-Si:H) with a thickness of 10–20 nm. The absorber i-layer with a thickness of 300 nm was made with an amorphous structure but close to the amorphous/nanocrystalline transition regime. Our standard device is on a stainless steel substrate. The device deposited on the GaAs(0 0 1) substrate is only for a flat cleavage of the device cross-sections, and has no electrical purpose. It is difficult to get flat cross-sectional surfaces on devices deposited on stainless steel substrates. The potential signal is measured from the long-range Coulomb force between the tip and sample. A rough surface morphology of the sample would result in a non-uniform Coulomb force if one included the Coulomb interaction between the side dimension of the tip and the sample surface. Therefore, a relatively flat surface is necessary for a reliable potential measurement. The several tens of nanometers of corrugation on the cross-sections of the device on GaAs substrates were adequately flat for the potential measurement. We also made sister samples on stainless steel substrates to ensure identical device performance with the different substrate materials.

953

1.5

1

0.5

0 0

1

2

3

4

5

6

7

-8

Bi Flux (×10 Torr) Fig. 1. Changes in the RMS roughness on the GaInNAs (Bi) surface with BEP of Bi evaporations, as measured by AFM.

measured by AFM. Without a Bi flux present, the surface roughness is 2 nm, which is the In-segregation-related surface roughening [13,14]. The epitaxial growth conditions with a substrate temperature of 530 1C are for a wide depletion region or low carrier concentration [12]. However, these conditions lead to the In segregation and rough surface. With a Bi evaporation in a beam equivalent pressure (BEP) of 0.4  108 Torr, the surface roughness dropped about one order of magnitude, to 0.3 nm. However, further increase of BEP in the 108 Torr range did not change the roughness significantly. Bi is considered to act as surfactant during the MBE growth, through increasing the diffusion length relative to the As-terminated surfaces and lowering the step energy by reducing adatom bond strength at the step edges [14]. Although increasing the Bi BEP in the 108 Torr range does not make the epitaxial layer smoother, it also induces a junction shift in the absorber layer, which leads to significant device degradation. Each potential profile in Fig. 2 is an average of 64 potential measurements on the sample with BEP of 0.4  108 Torr. We took the potential profiles under various external bias-voltages (Vb) to avoid the surface Fermi-level (EF) pinning effect. The profile at Vb ¼ 0 shows a 0.35 V potential drop at the p–n junction, which is much smaller than the bandgap of the absorber layer and is due to surface EF pinning [9,10,15]. SKPFM measures the surface potential. The surface depletions on both n- and p-regions make the surface potential contrast between the regions smaller than that in the bulk. However, if we apply a Vb on the device, the change in the surface potential should represent that in the bulk; in this way, we can avoid the surface EF pinning effect. This approach was published in our previous work involving junction studies [9,10,15]. The Vb-induced surface potential changes [Fig. 2(b)] are slightly smaller than the Vb values (0.9 V with Vb ¼ 1 V), which is due possibly to voltage losses on other device components such as the contacts and also to the long-range nature of Coulomb interaction in the measurements [16]. The potential change on the profile at the reverse bias Vb ¼ 0.25 V occurs over a length of

ARTICLE IN PRESS 954

C.-S. Jiang et al. / Ultramicroscopy 109 (2009) 952–957

n+-GaAs

2.0 mm, which is smaller than the depletion width of 2.5 mm as deduced from the C–V measurement. The shape of the potential profiles is also different from a one-dimensional simulation using the carrier concentration of 1.5  1014 cm3 that is deduced from the C–V measurement. The discrepancies between the SKPFM and C–V measurements are due possibly to the heavy asymmetry of the p–n junction, with 0.1-mm n-layer and 3-mm p-layer thicknesses. The much thicker p-layer makes the Coulomb interaction between the AFM tip and the sample, and thus the potential measurement, inclined toward the p-layer [16]. The Vb-induced electric fields [Fig. 2(c)] are further deduced from the Vb-induced potentials, and the location at the maximum electric field corresponds to the p–n junction. Note that regardless of the significant difference in the depletion widths and the potential profile shapes given by the SKPFM and C–V measurements, the maximum electric field locations under the various Vbs show consistency (within 40 nm) for Vb from 1.0 V through +1.0 V. This supports our approach of determining the p–n junction locations by taking the maximum locations of Vb-induced electric fields. We next show the direct observation of Bi-induced junction shifts. With increasing BEP to 0.67  108 Torr, location of the potential drop stays close to the front contact of the device [Fig. 3(a)]. However, the potential profile shape is significantly broadened in the absorber layer, and the maximum electric field location [Fig. 3(b)] is slightly shifted toward the back side. With further increasing BEP, both potential and electric field profiles exhibit clear shifts of the p–n junction to the back side of the absorber layer. The larger the BEP, the closer the junction is to the interface between the absorber and the substrate. Apparently, a conversion from p- to n-types occurred in the absorber layer due to the evaporation of Bi. Bi acts as surfactant and only a small amount of it is incorporated into the absorber layer. As reported in the previous article [12], no Bi was detected by SIMS above the background level (1016 cm3) on all the samples in Fig. 1. A small amount of Bi incorporation may be enough to change the doping configuration in the small background doping concentration of 1.5  1014 cm3. Admittance spectroscopy measurement has found a defect level associated with the Bi incorporation [12]. However, current–voltage measurements indicate an independency of dark saturation current on the Bi incorporation, possibly due to a much larger nitrogen-related background defect concentration that dominates the carrier transport across the junction [12]. The fact that the p–n junction location is closer to the interface of absorber/substrate with a higher Bi BEP suggests a build-up effect of the Bi incorporation and a gradual change in the donor concentration during MBE growth. It can be a build-up of Bi incorporation, defects, or impurities.

GaInNAs(Bi)

2.0 Vb =-1.0V -0.5

1.6

Potential (V)

-0.25 0

1.2

+0.25 +0.5 0.8

+1.0

0.4

0 Vb=-1.0 V -0.5 Potential difference (V)

0.8

-0.25 +0.25 +0.5

0.4

+1.0 0

-0.4

-0.8

Electric field (Arb. unit)

20

0 Vb=+1.0 V

-20

+0.5 +0.25

-40

4. Electric field uniformity in a-Si:H cells

-0.25 -0.5

-60

-1.0 -80 0

1

2

3

Distance (µm) Fig. 2. (a) SKPFM electrical potential, (b) the Vb-induced potential changes as deduced from (a), and (c) Vb-induced changes in the electric field as deduced from (b), taken on a cross-section of the GaInNAs (Bi) device with Bi BEP of 0.41 108 Torr. Positive and negative Vbs represent, respectively, the forward and reverse bias voltages.

In an ideal a-Si:H n–i–p junction, the potential should be linear and thus the electric field should be constant through the i-region of the junction. However, our SKPFM measurement shows that the electric field in the a-Si:H n–i–p devices is not uniform, but concentrates on both n/i and i/p interfaces. It further reveals a significant improvement of the electric field uniformity by depositing buffer layers at the interfaces. From the AFM image taken on the cross-section of an a-Si:H device [Fig. 4(a)], we are able to identify the indium-tin-oxide (ITO), a-Si:H, and Cr layers, as well as the GaAs substrate. The identification of the layers was further confirmed by using backscattering mode of scanning electron microscopy (SEM). Because the SKPM image [Fig. 4(b)] is quite uniform along the lateral

ARTICLE IN PRESS C.-S. Jiang et al. / Ultramicroscopy 109 (2009) 952–957

direction (parallel to the device surface), we averaged the SKPFM image along the lateral direction. The potential profile, which is given in Fig. 4(c), shows a valley around the Cr layer that results

n+-GaAs

GaAs buffer p-GaAs substrate

GaInNAs(Bi)

1000 6.1×10-8

Potential (mV)

900

800

2.7×10-8 1.1×10-8

700

600

500

0.67×10-8

PBi=0.41×10-8 Torr

Electric Field (arb. Unit)

400

1.1×10-8 2.7×10-8 6.1×10-8 0.67×10-8 PBi=0.41×10-8 Torr 0

1

2

3

4

5

6

7

Distance (µm) Fig. 3. (a) SKPFM electrical potential and (b) the electric field as deduced from (a), taken on the cross-sections of the GaInNAs(Bi) devices with various Bi BEP during MBE growth of the devices.

955

from the larger work function of Cr than both n-types a-Si:H and GaAs. However, the measured potential cannot be quantitatively understood by the bulk properties of the materials. Because the conduction band minima are nearly aligned between GaAs and a-Si:H [17,18], the potential should be about the same at both sides of the potential valley. However, the potential profile shows a large difference (380 mV) between the n-types a-Si:H and GaAs. This is due partially to the thin n-type a-Si:H layer thickness (20–30 nm); the long-range nature of Coulomb force makes the potential measurement on this thin a-Si:H layer smaller than its actual value [16]. Another reason for the potential difference is surface EF pinning. Different EF pinning positions possibly make the surface potentials on the n-type a-Si:H layer and GaAs substrate different. The potential decrease in the junction region (450 mV) is much smaller than the built-in potential (1.2 V) [19], which could be due to the surface EF pinning and the thinness (20–30 nm) of both n- and p-layers. Similar to the case of GaInNAs(Bi), we applied Vb to the a-Si:H device to avoid the surface EF pinning effects. The potential profiles under various Vbs and the Vb-induced potential changes are shown in Fig. 5(a) and (b). The values of potential changes are consistent with the corresponding Vbs, indicating that the surface potential change is identical to that in the bulk, which supports our approach of taking the Vb-induced potential change. The Vb-induced change in the electric field is further deduced [Fig. 5(c)]. One sees that the electric field does not distribute uniformly through the n–i–p structure, but concentrates on both n/i and i/p interfaces. This electric filed distribution indicates higher charge densities (traps or defects) at the n/i and i/p interfaces than in the middle of the i-layer, which is consistent with the simulations [20]. The electric field extensions into both front and back contact layers are due most likely to the spatial resolution of the SKPFM technique. The potential measurements shown in Figs. 5 and 6(a) are taken on the a-Si:H sample with buffer layers at both n/i and i/p interfaces. The 10–20-nm-thick buffer layers were a-Si:H made at lower substrate temperature than the i-layer deposition. The low substrate temperature increased the bandgap, which reduced carrier back diffusion and improved the electric field distribution and solar cell performance. We also measured the potential profiles in an a-Si:H sample without the buffer layers [Fig. 6(b)]. Comparison of the Vb-induced electric field profiles leads to the observation that the electric field is more non-uniform in the cell with no buffer layers [Fig. 6(b)] than in the one with the buffer layers [Fig. 6(a)]. The higher Vb-induced electric field at the n/i and i/p interfaces indicates higher charge redistributions under the electrical field, due to higher defect concentrations at the interfaces. The incorporation of the buffer layers effectively

Potential (V)

ITO p

800

i Si

600

n Cr

400

GaAs

200 nm

200

Distance(nm)

0.4

AFM

0 Fig. 4. (a) An AFM and (b) the corresponding SKPFM images taken on a cross section of the a-Si:H n–i–p device deposited on a GaAs(0 0 1) substrate. (c) displays a potential profile that is an average of the SKPFM image along the lateral direction.

ARTICLE IN PRESS 956

C.-S. Jiang et al. / Ultramicroscopy 109 (2009) 952–957

a-Si GaAs

Cr

n

i

a-Si p

ITO

Cr

n

200

400 Distance (nm)

GaAs

i

p

ITO

Vb =+1.0 V +0.8 +0.6

Electric field (arb. Unit)

0

Potential (V)

+0.4

+0.2 0 -0.2 V -0.4 0.8

Vb=+0.2 V +0.4

-0.6

+0.6

-0.8

+0.8 +1.0

-1.0

0 Vb =+1.0 V +0.8

GaAs

Cr

n

600 a-Si i

p

ITO

Potential Difference (V)

+0.6 +0.4

Electric field (arb. unit)

+0.2

Vb =-0.2 V -0.4 -0.6 -0.8

0.8

-1.0

0 Vb=+0.2 V +0.4 +0.6 +0.8

Vb =-1.0 V

+1.0

-0.8 Electric field (arb. Unit)

-0.6

0

-0.4 -0.2

200

400 600 Distance (nm)

800

Fig. 6. Vb-induced changes in the electric field on the junction of a a-Si:H device (a) with and (b) without the buffer layers at both the n/i and i/p interfaces.

0 reduced the non-uniformity of the electric field distribution by reducing the trapped charges at the interfaces, which further improved the device performance parameters of open-circuit voltage, fill factor, and energy conversion efficiency.

Vb =+0.2 +0.4 +0.6 +0.8

5. Summary

+1.0 0

200

400

600

Distance (nm) Fig. 5. (a) SKPFM electrical potential, (b) the Vb-induced potential changes as deduced from (a), and (c) Vb-induced changes in the electric field as deduced from (b), taken on a cross-section of the a-Si:H device. Positive and negative Vbs represent, respectively, the forward and reverse bias voltages.

We presented two examples of microelectrical characterizations in junctions of solar cell devices, using SKPFM. In the first characterization, Bi coevaporation during MBE growth of GaInNAs made the epitaxial layer smoother, acting as a surfactant, but it induced a shift of the p–n junction to the back side of the absorber layer, and thus a serious device degradation. In the second characterization, the Vb-induced electric field exhibited high nonuniformity across the n–i–p junction of the a-Si:H devices, due possibly to high defect concentrations at the n/i and i/p interfaces.

ARTICLE IN PRESS C.-S. Jiang et al. / Ultramicroscopy 109 (2009) 952–957

The electric field uniformity was significantly improved by depositing buffer layers at the interfaces.

Acknowledgements This work was supported by the US Department of Energy under contract number DE-AC36-99GO10337 at NREL and by DOE under the Solar America Initiative Program Contract no. DE-FC3607 GO 17053 at United Solar Ovonic LLC. References [1] E. Schmich, H. Lautenschager, T. Frieb, F. Trenkle, N. Schillinger, S. Reber, Prog. Photovolt.: Res. Appl. 16 (2008) 159. [2] A. Bentzen, B.G. Svensson, E.S. Marstein, A. Holt, Solar Energy Mater. & Solar Cells 90 (2006) 3193. [3] S.E. Asher, K. Ramanathan, D.W. Niles, H. Moutinho, In: Proceedings of the 15th NCPV Photovoltaic Program Review Meeting, Denver, CO, USA, September 1998, p. 126. [4] S.S. Hegedus, W.N. Shafarman, Prog. Photovolt.: Res. Appl. 12 (2004) 155. [5] J.P. Kleider, A.S. Gudovskikh, Mat. Res. Soc. Symp. Proc. 1066 (2008) A4.1.

957

[6] M. Nonnenmacher, M.P. O’Boyle, H.K. Wickramasinghe, Appl. Phys. Lett. 58 (1991) 2921. [7] A. Kikukawa, S. Hosaka, R. Imura, Appl. Phys. Lett. 66 (1995) 3510. [8] S. Ono, T. Takahashi, Jpn. J. Appl. Phys. 44 (2005) 6213. [9] C.-S. Jiang, H.R. Moutinho, D.J. Friedman, J.F. Geisz, M.M. Al-Jassim, J. Appl. Phys. 93 (2003) 10035. [10] C.-S. Jiang, F.S. Hasoon, H.R. Moutinho, H.A. Al-Thani, M.J. Romero, M.M. Al-Jassim, Appl. Phys. Lett. 82 (2003) 127. [11] S. Saraf, A. Schwarzman, Y. Dvash, S. Cohen, D. Ritter, Y. Rosenwaks, Phys. Rev. B73 (2006) 035336. [12] A.J. Ptak, R. France, C.-S. Jiang, R.C. Reedy, J. Vac. Sci. Technol. B26 (2008) 1053. [13] M.R. Pillai, S.-S. Kim, S.T. Ho, S.A. Barnett, J. Vac. Sci. Technol. B18 (2000) 1232. [14] S. Tixier, M. Adamcyk, E.C. Young, J.H. Schmid, T. Tiedje, J. Cryst. Growth 251 (2003) 449. [15] C.-S. Jiang, D.J. Friedman, J.F. Geisz, H.R. Moutinho, M.J. Romero, M.M. Al-Jassim, Appl. Phys. Lett. 83 (2003) 1572. [16] Y. Shen, D.M. Barnett, P.M. Pinsky, Rev. Sci. Instrum. 79 (2008) 023711. [17] X. Xu, J. Yang, A. Banerjee, S. Guha, K. Vasanth, S. Wagner, Appl. Phys. Lett. 67 (1995) 2323. [18] M.E. Levinshtein, S.L. Rumyantsev, in: M. Levinshtein, S. Rumyantsev, M. Shur (Eds.), Handbook Series on Semiconductor Parameters, vol. 1, World Scientific, River Edge, NJ, 1996. [19] L. Jiang, J.H. Lyou, S. Rane, E.A. Schiff, Q. Wang, Q. Yuan, Mater. Res. Soc. Symp. Proc. 609 (2000) A18.3. [20] M. Hack, M. Shur, J. Appl. Phys. 58 (1985) 997.