Solid State Communications 144 (2007) 153–157 www.elsevier.com/locate/ssc Metal–Insulator Transition and superconductivity in doped semiconductors A. Therese Pushpam a,∗ , K. Navaneethakrishnan b a Fatima College, Madurai, 625 018, India b School of Physics, Madurai Kamaraj University, Madurai, 625 021, India Received 9 April 2007; received in revised form 23 July 2007; accepted 26 July 2007 by F. Peeters Available online 8 August 2007 Abstract Using the effective mass theory and the variational method, the critical concentrations of shallow donors in Si at which the Mott Transition occurs are obtained. Our theory uses the experimentally available donor ionization energies in the low-concentration regime and a mass variation with interimpurity distance to account for the impurity bands. Since the experimentally available donor ionization energies are used, the centralcell effects and valley–orbit interactions are taken into account. Excellent agreement with the available experimental results is obtained for P, As, Sb and Bi donors in Si. Since the critical concentration at which superconductivity occurs in B doped Si is established recently, we predict the concentrations at which superconductivity should occur for these donors also. c 2007 Elsevier Ltd. All rights reserved. PACS: 71.55-i; 71.55.Cn; 71.55Eq; 71.55Gs; 71.30.+h; 71.70.Di Keywords: A. Semiconductors; A. Doped semiconductors; D. Metal–Insulator Transition; D. Superconductivity 1. Introduction Metal–Insulator Transition in semiconductors is drawing considerable attention at present for several reasons. The problem of impurity bands and their role in MIT and superconductivity have not been completely understood [1– 5]. The predictions in the Hubbard model for the physically interesting situation ( Ut ≈ 1) have eluded confirmations by exact solution. Also contrary to the one electron scaling theory [6], MIT has been observed in several low-dimensional semiconducting systems throwing a challenge to the theory [7]. Several recent works have predicted superconductivity in semiconductors and in alkali halide systems under hydrostatic pressure [8]. Many of these results have been confirmed experimentally. MIT has been shown to be the precursor for superconductivity not only in high Tc superconductors, but also in several other systems including doped semiconductors [3, 9]. The co-existence of ferromagnetism and superconductivity and MIT in intense magnetic fields are other issues of recent interest [10–13]. ∗ Corresponding author. E-mail address: tpushpu@yahoo.co.in (A. Therese Pushpam). 0038-1098/$ - see front matter c 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ssc.2007.07.036 Many-valley semiconductors pose considerable interest not only in the formulation of effective mass theory but also in the prediction of superconductivity for a long time [14–16] Even today, no effective mass theory exists for these systems that is free from criticism. In direct-gap semiconductors such as GaAs, the simple effective mass theory which pictures the donor system as a hydrogenic problem with an effective mass for the electron pertinent to the conduction band minimum, and the potential energy reduced by static dielectric constant works well [17]. Several drawbacks of the effective mass theory (EMT) as applied to a many-valley semiconductor such as Si are attributed to central-cell corrections [18]. Because of these deficiencies, the predictions of the critical concentrations at which MIT occurs in these systems differ from experimental values by one or two orders [19,20]. Other difficulties in the theoretical formulations include randomness in impurity distribution which triggers Anderson Transition and e–e correlations leading to the formation of impurity bands [1,6]. In the present communication, we present a one electron theory within the effective mass formulation using effective masses obtained from the experimental, low-concentration donor ionization energies, describing a mass variation with the impurity concentrations and providing critical donor 154 A. Therese Pushpam, K. Navaneethakrishnan / Solid State Communications 144 (2007) 153–157 Table 1 Ionization energies and effective masses for donors in Si Donor P As Sb Bi Table 2 Critical donor concentrations for MIT using Eq. (4) E ion (meV) experimenta m ∗ (a.u.) experiment m ∗ (a.u.) average 45.47 53.69 42.68 70.90 0.4577 0.5404 0.4296 0.7136 0.378 0.419 0.364 0.506 a Ref. [21]. concentrations at which MIT occurs for several donors in Si. Using the recently observed superconductivity for B doped Si, we predict the concentrations at which superconductivity should be expected for P, As, Sb and Bi donors in Si. 2. MIT in n-type Si For low-donor concentrations (≤1014 cm−3 ), the effective mass theory in its simplest form leads to the solution of the Schrödinger equation " # − h̄ 2 2 e2 ∇ − ψ(r ) = Eψ(r ), (1) 2m ∗ εo r where m ∗ is the effective mass at the conduction band minimum, which in the spherical band approximation is 0.2982 a.u. and εo is the static dielectric constant which is 11.7 for Si. ψ(r ) is called an envelope function [15]. To bring in the impurity dependence into the theory, we use the experimentally available donor ionization energies for the effective mass. These values are given in Table 1. The usage of these effective masses for different donors takes into account all central-cell effects and the valley–orbit interactions. In the present work, we are only interested in MIT and subsequent superconductivity and not in the valley–orbit split excited states [21]. When the impurity concentration increases, overlap effects become appreciable leading to the formation of impurity bands. MIT occurs when the Hubbard gap closes as concentration increases [1,4]. It has been recently shown in a series of works that this simple one electron picture leading to MIT is consistent with the Hubbard model calculations when concentrationdependent functions in the potential energy term such as Thomas–Fermi and Hartree–Fock functions are used [22]. The properties of the impurity bands are not completely understood though considerable efforts are spent. Since the donor electron enters into the impurity band when metallization occurs, we use an effective mass which is the average of the low-concentration impurity-dependent effective mass and 0.2982. A mass of 0.2982 for donors in Si is obtained by equating the donor ionization energy in the spherical band approximation to the numerical value obtained by using the longitudinal and the transverse effective masses (m t and m l ) pertinent to the mass ellipsoid near the conduction band minimum [23]. See the foot note of Table 2. These masses for P, As, Sb and Bi donors in Si are 0.378, 0.419, 0.364 and 0.506 respectively. Only when the impurity band merges with the conduction band, the use of the mass 0.2982 is justified. Using the Thomas–Fermi screening n c in cm−3 Present m ∗ = 0.2982 Semiconductor Presenta Experiment 1.74 × 1018 2.86 × 1018 1.43 × 1018 6.58 × 1018 3.85 × 1015 3.525 × 1018 3.74 × 1018 18 4.815 × 10 9 × 1018 3.147 × 1018 3.5 × 1018 8.458 × 1018 – – 1.5 × 1016 hm i exp t +0.2982 a Corresponds to the concentration for m . average = 2 Si (P) Si (As) Si (Sb) Si (Bi) GaAs function ε(r1 ) = 1ε e−r λ , we write the Hamiltonian as e2 h̄ 2 k 2 − e−λr , ∗ 2m rε where (2) H= λ= " 12π e2 m ∗ h̄ 2 ε # 12 1 −2 n 3 . [3π 2 ] 3 γ (3) Here n is the electron concentration and γ is the number of equivalent conduction band minima (6 in Si and 1 in GaAs) in a semiconductor. Since the Schrödinger equation for H is not amenable to an exact solution, we try an alternative approach which has its origin in atomic physics. The stationary state is characterized by an integral multiple of de Broglie waves filling the circumference of the impurity orbital. Hence 2πan = nλ and k = ann . For the ground state, n = 1. Replacing r by a and minimizing H with respect to a leads to ln a ∗ = ln a + ln(1 + aλ) − aλ, (4) εo h̄ where a ∗ = m ∗ e2 is the effective-Bohr radius. As in Ref. [24] we have obtained the effective mass as a function of interimpurity distance using values between m exp t and 0.2982. This is shown in Fig. 1 for As in Si. We note that when n c ≈ 1018 cm−3 , we obtain m ∗ ≈ 0.42 for As donor. Eq. (4) is an exact result. More sophisticated Hamiltonians including overlap and correlations lead either to approximate results or numerical procedures [5]. Eq. (4) is solved for different impurity concentrations. When these values of a are used in Eq. (2), we obtain donor binding energies. The results are presented in Fig. 2 for As donor in Si. For other donors, the critical concentrations are presented in Table 2. If the relativistic correction is used, then " #1/2 h̄ 2 n 2 e2 2 H = mc 1 + 2 2 2 (5) − e−λr rε m c r 2 and we obtain " # 1 h̄ 2 ln a = ln a + ln(1 + aλ) − aλ − ln 1 + 2 2 2 2 m c a ∗ 2 (6) in place of Eq. (3). However, since m 2h̄c2 a 2 ≺≺ 1, Eq. (5) leads to the same result for ‘a’ obtained earlier from Eq. (4). A. Therese Pushpam, K. Navaneethakrishnan / Solid State Communications 144 (2007) 153–157 155 Fig. 1. Variation of effective mass with interimpurity distance for As in Si. Fig. 3. Polarizability versus concentration for As in Si. Fig. 4. Variation of diamagnetic susceptibility with concentration for As in Si. Fig. 2. Donor ionization energy as a function of As donor concentration in Si. Hence we find no change either in the ionization energies or in the critical concentration due to relativistic effects. This is physically due to the fact that in semiconductors, the effective Bohr radii are very large and the binding energies are weaker. Hence, relativistic corrections appear smaller. The variation of donor polarizability α D with impurity concentration is also of great interest for a long time [25]. At the critical concentration, the donor polarizability has been observed to diverge indicating nonlocalization of the carrier [26]. In our model, α D = 4.5a 3 [27]. Hence, we have plotted in Fig. 3, the variation of α D with concentration. The rapid variation at n c is evident. The diamagnetic susceptibility is given by χdia = 2 − 6mce 2 ε hr 2 i for a single donor [28]. The variation of χdia with o concentration is shown in Fig. 4. The divergence at n c is again clear from this figure. Similar works have appeared recently on low-dimensional semiconductor systems and our results are in agreement with these results in the proper limit [29]. Whether the MIT takes place in the conduction band or in the impurity band is a highly debatable question. Whether an impurity band can support metallic conduction is also not clearly established. However, by observing the 1s → 2p transition of a shallow donor, Romero et al. [30] have concluded that the impurity band merges with the conduction band of GaAs for a donor concentration of 8 × 1016 cm−3 . Similar conclusion in a GaAs/Ga1−x Alx As superlattice has been arrived at by Helm et al. [31]. In contrast, it has been shown that the Hubbard model results are in agreement with the mean-field calculations (Thomas–Fermi or Hartree) as far as the critical concentration values are concerned if proper corrections are made for the effective masses due to overlap effects [22,26]. 3. Superconductivity Superconductivity in semiconductors is not new. Traditionally there have been two approaches to superconductivity in semiconductors: (i) In undoped samples, by applying hydrostatic pressure, closure of the band gap leading to metallization and subsequent increase in pressure enhancing the density of states at the Fermi level leading to BCS like superconductivity [32]. The Tc values observed are low (≈3 K). (ii) By increasing the impurity concentration, the formation of impurity bands and the merging of the impurity band with the conduction band leading to superconductivity [33,34]. Since the conduction in the impurity band is by hopping, and in view of the recent experimental evidence for superconductivity in B doped 156 A. Therese Pushpam, K. Navaneethakrishnan / Solid State Communications 144 (2007) 153–157 Si at n of the order of 1021 cm−3 , it is clear that superconductivity occurs in the valence band due to holes with the conventional BCS-type mechanism [9]. Further experimental support for BCS mechanism in the conventional band structure of Si follows from similar observations in B doped diamond [35]. From the available theoretical and experimental works on pressure induced superconductivity on several metals, alkali halides and semiconducting systems, it appears to us that the high frequency, long wave vector phonons are eliminated encouraging conduction electrons to undergo Cooper pairing in the presence of low energy and long wavelength phonons. At high pressures the following effects are expected. (i) The band gap closes leading to metallization. (ii) Increase in Fermi energy and the consequent increase in the density of electron states. (iii) Enlargement of the Brillouin zone associated with increase in phonon energies including Debye energy and the consequent increase in phonon density of states. However, at low temperatures high-energy phonons are not excited and superconductivity occurs by BCS mechanism. In doped systems however, high impurity concentrations are required to attain the superconducting state. From studies such 1 as B doped Si [9], approximately 10 th of the host atoms are replaced by impurity atoms. In such a situation, the phonon dispersion relations of the host material require enormous modification in the discussion of superconductivity in the BCS theory. Such a problem has not been addressed so far. Roughly speaking, in an impurity atom with a heavier mass such as V group donors in Si, one expects a reduction in the phonon frequencies in addition to the defect modes. Since the impurity band merges with the conduction band when superconductivity occurs [9,35], the consequent increase in Fermi energy and the elimination of large momentum and large energy phonons from the scenario at low temperatures leads to superconductivity as in the pressure applied case. Using the experimentally available donor concentrations in Si, for the onset of super conductivity, we predict the following concentrations for P, As, Sb and Bi donors in Si. Since the ionization energies vanish when MIT occurs, we use the experimentally observed critical concentration of 2.8 × 1021 cm−3 for B in Si, and the critical donor concentration ion (B) for P, As, Sb and Bi should be EEion (D) n c (B), where D refers to any donor and E ion (B) is 45 meV. Since Bi is the heaviest atom, we predict donor concentrations of 2.77 × 1021 cm−3 , 2.35 × 1021 cm−3 , 2.95 × 1021 cm−3 and 1.78 × 1021 cm−3 for P, As, Sb and Bi donors respectively. In an earlier work, Rasolt predicted superconductivity in many-valley semiconductors in the presence of intense magnetic fields [16]. Typical values obtained by Rasolt for n are of the order of 1018 cm−3 and Tc of the order of 0.5 K for a magnetic field of 1 MG. Recently, it has been shown that before entering into the superconductivity state the doped semiconductor passes through a metallic state in every magnetic field [36]. In view of recent discovery of superconductivity in ferromagnetic systems, we believe that it is worthwhile to study experimentally the magnetic field induced superconductivity. In an earlier work Cohen has shown that Tc values enhance when the number of conduction band valleys increase [15]. In the last few years, superconductivity in B doped Diamond has drawn the attention of several physicists [9]. The results of Ref. [36] may further be improved when the appropriate screening function relevant to the host semiconductor is used. 4. Results and discussion The critical donor concentrations at which metallization occurs is presented in Table 2 for different donors in Si. We find an excellent agreement with experimental values. Such an agreement obtained for the first time confirms that metallization occurs in the impurity band since the mass at n c is different from the mass at the conduction band minimum, namely 0.2982. This observation is in agreement with several experiments performed recently indicating the role of impurity bands [37]. It also demonstrates that the correlation effects among electrons, the central-cell effects and valley–orbit interactions on the donor states are correctly taken in to account. Sharp increase in the effective mass near the critical density has also been suggested recently by Shaskin et al. [38]. This suggestion is in agreement with the masses used in the present work, displayed in Table 1. We believe that this increase in the mass from the conduction band minimum value of 0.2982 is due to impurity bands. The prediction of donor concentration above which superconductivity should occur in doped Si requires experimental support. The concentrations predicted are 2 to 3 orders higher than what Rasolt conjectured in a strong magnetic field [16]. Cohen earlier predicted the donor concentration of the order of 1020 cm−3 in many-valley semiconductors [15]. However, the recently observed impurity concentrations in Si are of the order of 1021 cm−3 . We conclude that though the Tc values observed so far for any semiconductor (except High Tc materials) is less than 10 K, these systems offer a scope for understanding impurity bands, MIT and the mechanism for superconductivity. Since nanotechnology involving semiconductors is expected to give enormous impetus to technology, understanding the phenomena in these systems will also be of great interest. Hence MIT and superconductivity in lower dimensions should also be pursued with great interest. References [1] N.F. Mott, Metal–Insulator Transition, Taylor & Francis Ltd., London, 1974. [2] R.N. Bhatt, Physica B 146 (1987) 99. [3] P.W. Anderson, Theory of Superconductivity in Cuprates, Princeton University Press, 1997. [4] H. Kamimura, in: P.N. Butcher, N.H. March, M.P. Tosi (Eds.), Crystalline Semiconducting Materials and Devices, Plenum Press, New York, 1986. [5] Masatoshi Imada, Atsushi Fujimori, Yshinori Tokura, Rev. Modern Phys. 70 (1996) 1039. [6] P.A. Lee, T.V. Ramakrishnan, Rev. Modern Phys. 57 (1985) 287 and references therein. [7] E. Abrahams, S.V. Kravchenko, M.P. Sarachik, Rev. Modern Phys. 73 (2001) 251 and references therein. [8] K. Iyakutti, et al., J. Phys.: Condens. Matter. 16 (2004) 1. [9] E. Bustarret, et al., Nature (London) 444 (2006) 465. A. Therese Pushpam, K. Navaneethakrishnan / Solid State Communications 144 (2007) 153–157 [10] K. Shimizu, et al., Nature (London) 412 (2001) 316. [11] S.S. Saxena, et al., Nature (London) 406 (2001) 587. [12] C. Pfleiderer, et al., Nature (London) 412 (2001) 58. [13] D. Aoki, Nature (London) 413 (2001) 613. [14] R. Resta, in: P.N. Butcher, N.H. March, M.P. Tosi (Eds.), Crystalline Semiconducting Materials and Devices, Plenum Press, new York, 1986, p. 217. [15] M.L. Cohen, Phys. Rev. 134 (1964) A511. [16] M. Rasolt, Phys. Rev. Lett. 58 (1987) 1482. [17] W. Kohn, Solid State Phys. 5 (1957) 257. [18] Sr. Gerardin Jayam, K. Navaneethakrishnan, J. Appl. Phys. 89 (2001) 6198. [19] A. John Peter, K. Navaneethakrishnan, Phys. Status Solidi (b) 220 (2000) 897. [20] M.N. Alexander, D.F. Holcomb, Rev. Modern Phys. 40 (1968) 815. [21] A.K. Ramdas, S. Rodriguez, Rep. Prog. Phys. 44 (1981) 1297. [22] A. John Peter, K. Navaneethakrishnan, Physica E 16 (2003) 223–229. 157 [23] T.H. Ning, C.T. Sah, Phys. Rev. B 4 (1971) 3468. [24] A. John Peter, K. Navaneethakrishnan, Physica E 15 (2002) 153. [25] M. Capizzi, et al., Phys. Rev. Lett. 44 (1980) 1019. [26] A. John Peter, K. Navaneethakrishnan, Physica B 366 (2005) 146. [27] L.I. Schiff, Quantum Mechanics, third ed., McGraw Hill, 1968, p. 265. [28] C. Kittel, Introduction to Solid State Physics, seventh ed., John Wiley, 1996, p. 418. [29] P. Nithananthi, K. Jayakumar, Solid State Commun. 138 (2006) 305; Superlatt. Microstruct. 40 (2006) 174. [30] D. Romero, et al., Phys. Rev. B 42 (1990-I) 3179. [31] M. Helm, et al., Phys. Rev. B 48 (1993-I) 1601. [32] S.B. Zhang, et al., Solid State Commun. 71 (1989) 369. [33] Y.K. Li, et al., J. Appl. Phys. 71 (1992) 2018. [34] G. Baskaran, Preprint at http://arXiv.org/cond-mat/0404286, 2004. [35] T. Yokova, et al., Nature (London) 438 (2005) 647. [36] R. Suganya, K. Navaneethakrishnan, Solid State Commun. 138 (2006) 99. [37] W. Hilber, et al., Phys. Rev. B 53 (1996-I) 6919. [38] A. Shashkin, et al., Phys. Rev. B 66 (2002) 073303.
0
You can add this document to your study collection(s)
Sign in Available only to authorized usersYou can add this document to your saved list
Sign in Available only to authorized users(For complaints, use another form )