Skip to main content
Log in

Asymptotic Linear Spectral Statistics for Spiked Hermitian Random Matrices

  • Published:
Journal of Statistical Physics Aims and scope Submit manuscript

Abstract

Using the Coulomb Fluid method, this paper derives central limit theorems (CLTs) for linear spectral statistics of three “spiked” Hermitian random matrix ensembles. These include Johnstone’s spiked model (i.e., central Wishart with spiked correlation), non-central Wishart with rank-one non-centrality, and a related class of non-central \(F\) matrices. For a generic linear statistic, we derive simple and explicit CLT expressions as the matrix dimensions grow large. For all three ensembles under consideration, we find that the primary effect of the spike is to introduce an \(O(1)\) correction term to the asymptotic mean of the linear spectral statistic, which we characterize with simple formulas. The utility of our proposed framework is demonstrated through application to three different linear statistics problems: the classical likelihood ratio test for a population covariance, the capacity analysis of multi-antenna wireless communication systems with a line-of-sight transmission path, and a classical multiple sample significance testing problem.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

Notes

  1. The paper [35] considered a more general model than the spiked model. Therein, CLT results were presented for linear spectral statistics, with the key quantities involving solutions to implicit equations.

  2. We point out that [34] (see also [60]) considered a generalized problem which allowed for multiple spiked eigenvalues.

  3. Note that \(x\text { dB}=10\log _{10}x\). These are the typical units used for expressing SNR in wireless communication systems.

  4. Note that under \(H_0\), \(\bar{\mu }_{\text {R}} = 0\).

  5. That is, when performing the test, the same decision threshold was chosen, as based on the asymptotic Gaussian distribution under \(H_0\).

References

  1. Ledoit, O., Wolf, M.: Honey, I shrunk the sample covariance matrix. J. Portf. Manag. 30(4), 110–119 (2004)

    Article  MathSciNet  Google Scholar 

  2. Rubio, F., Mestre, X., Palomar, D.P.: Performance analysis and optimal selection of large minimum variance portfolios under estimation risk. IEEE J. Sel. Topics Signal Process. 6(4), 337–350 (2012)

    Article  ADS  Google Scholar 

  3. Dahirel, V., Shekhar, K., Pereyra, F., Miura, T., Artyomov, M., Talsania, S., Allen, T.M., Altfeld, M., Carrington, M., Irvine, D.J., Walker, B.D., Chakraborty, A.K.: Coordinate linkage of HIV evolution reveals regions of immunological vulnerability. Proc. Natl. Acad. Sci. 108(28), 11530–11535 (2011)

    Article  ADS  Google Scholar 

  4. Quadeer, A.A., Louie, R.H., Shekhar, K., Chakraborty, A.K., Hsing, I., McKay, M.R.: Statistical linkage analysis of substitutions in patient-derived sequences of genotype 1a Hepatitis C virus non-structural protein 3 exposes targets for immunogen design. J. Virol. 88(13), 7628–7644 (2014)

    Article  Google Scholar 

  5. Hoydis, J., ten Brink, S., Debbah, M.: Massive mimo in the ul/dl of cellular networks: how many antennas do we need? IEEE J. Sel. Areas Commun. 2(2), 160–171 (2013)

    Article  Google Scholar 

  6. Matthaiou, M., Zhong, C., McKay, M.R., Ratnarajah, T.: Sum rate analysis of zf receivers in distributed mimo systems. IEEE J. Sel. Areas Commun. 2(2), 180–191 (2013)

    Article  Google Scholar 

  7. Dyson, F.J., Mehta, M.L.: Statistical theory of the energy levels of complex systems. J. Math. Phys. 4(5), 701–712 (1963)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  8. Beenakker, C.: Universality in the random-matrix theory of quantum transport. Phys. Rev. Lett. 70(8), 1155–1158 (1993)

    Article  ADS  Google Scholar 

  9. Beenakker, C.: Random-matrix theory of mesoscopic fluctuations in conductors and superconductors. Phys. Rev. B 47(23), 15763–15775 (1993)

    Article  ADS  Google Scholar 

  10. Beenakker, C.: Universilaty of Br Tzin and Zee’s spectral correlator. Nucl. Phys. B 422, 515–520 (1995)

    Article  ADS  Google Scholar 

  11. Chen, Y., Lawrence, N.: On the linear statistics of Hermitian random matrices. J. Phys. A 31(4), 1141–1152 (1998)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  12. Lytova, A., Pastur, L.: Central limit theorem for linear eigenvalue statistics of random matrices with independent entries. Ann. Probab. 37(5), 1778–1840 (2009)

    Article  MATH  MathSciNet  Google Scholar 

  13. Costin, O., Lebowitz, J.: Gaussian fluctuation in random matrices. Phys. Rev. Lett. 75(1), 69–73 (1995)

    Article  MathSciNet  ADS  Google Scholar 

  14. Diaconis, P., Evans, S.N.: Linear functional of eigenvalues of random matrices. Trans. Am. Math. Soc. 353(7), 2615–2633 (2001)

    Article  MATH  MathSciNet  Google Scholar 

  15. Zheng, S.: Central limit theorems for linear spectral statistics of large dimensional \(F\)-matrices. Ann. Inst. Henri Poincaré Probab. Stat. 48(2), 444–476 (2012)

    Article  MATH  ADS  Google Scholar 

  16. Johansson, K.: On fluctuations of eigenvalues of random Hermitian matrices. Duke Math. J. 91, 151–204 (1998)

    Article  MATH  MathSciNet  Google Scholar 

  17. Bai, Z.D., Jiang, D., Yao, J.-F., Zheng, S.: Corrections to LRT on large-dimensional covariance matrix by RMT. Ann. Stat. 37(6B), 3822–3840 (2009)

    Article  MATH  MathSciNet  Google Scholar 

  18. Chen, Y., McKay, M.R.: Coulumb fluid, Painlevé transcendents, and the information theory of MIMO systems. IEEE Trans. Inform. Theory 58(7), 4594–4634 (2012)

    Article  MathSciNet  Google Scholar 

  19. Bai, Z.D., Jiang, D., Yao, J.-F., Zheng, S.: Testing linear hypotheses in high-dimensional regressions. Statistics 47(6), 1207–1223 (2013)

    Article  MATH  MathSciNet  Google Scholar 

  20. Anderson, G.W., Zeitouni, O.: A CLT for a band matrix model. Probab. Theory Relat. Fields 134(2), 283–338 (2005)

    Article  MathSciNet  Google Scholar 

  21. Johnstone, I.M.: On the distribution of the largest eigenvalue in principal components analysis. Ann. Stat. 29(2), 295–327 (2001)

    Article  MATH  MathSciNet  Google Scholar 

  22. Baik, J., Ben, G.: Arous, S. Péché, Phase transition of the largest eigenvalue for non null complex sample covariance matrices. Ann. Probab. 33(5), 1643–1697 (2005)

    Article  MATH  MathSciNet  Google Scholar 

  23. Baik, J., Silverstein, J.: Eigenvalues of large sample covariance matrices of spiked population models. J. Multivar. Anal. 97(6), 1382–1408 (2006)

    Article  MATH  MathSciNet  Google Scholar 

  24. Paul, D.: Asymptotics of sample eigenstructure for a large dimensional spiked covariance model. Stat. Sinica. 17(4), 1617–1642 (2007)

    MATH  Google Scholar 

  25. Bai, Z.D., Yao, J.-F.: Central limit theorems for eigenvalues in a spiked population model. Ann. Inst. Henri Poincaré Probab. Stat. 44(3), 447–474 (2008)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  26. Rao Nadakuditi, R., Silverstein, J.W.: Fundamental limit of sample generalized eigenvalue based detection of signals in noise using relatively few signal-bearing and noise-only samples. IEEE J. Sel. Top. Signal Proc. 4(3), 468–480 (2010)

    Article  Google Scholar 

  27. Bai, Z.D., Yao, J.-F.: On sample eigenvalues in a generalized spiked population model. J. Multivar. Anal. 106, 167–177 (2012)

    Article  MATH  MathSciNet  Google Scholar 

  28. Rao Nadakuditi, R., Edelman, A.: Sample eigenvalue based detection of high-dimensional signals in white noise using relatively few samples. IEEE Trans. Signal Process. 56(7), 2625–2638 (2008)

    Article  MathSciNet  ADS  Google Scholar 

  29. Torun, M.U., Akansu, A.N., Avellaneda, M.: Portfolio risk in multiple frequencies. IEEE Signal Process. Mag. 5(28), 61–71 (2011)

    Article  Google Scholar 

  30. Couillet, R., Debbah, M.: Random Matrix Methods for Wireless Communications, 1st edn. Cambridge University Press, New York (2011)

    Book  MATH  Google Scholar 

  31. Bianchi, P., Debbah, M., Maida, M., Najim, J.: Performance of statistical tests for single-source detection using random matrix theory. IEEE Trans. Inform. Theory 57(4), 2400–2419 (2011)

    Article  MathSciNet  Google Scholar 

  32. Couillet, R., Hachem, W.: Fluctuations of spiked random matrix models and failure diagnosis in sensor networks. IEEE Trans. Inform. Theory 59(1), 509–525 (2013)

    Article  MathSciNet  Google Scholar 

  33. Simon, S.H., Moustakas, A.L., Marinelli, L.: Capacity and character expansions: moment-generating function and other exact results for MIMO correlated channels. IEEE Trans. Inform. Theory 53(12), 5336–5351 (2006)

    Article  MathSciNet  Google Scholar 

  34. Wang, Q., Silverstein, J.W., Yao, J.-F.: A note on the CLT of the LSS for sample covariance matrix from a spiked population model. J. Multivar. Anal. 130, 194–207 (2014)

    Article  MATH  MathSciNet  Google Scholar 

  35. Bai, Z.D., Silverstein, J.W.: CLT for linear spectral statistics of large-dimensional sample covariance matrices. Ann. Probab. 32(1A), 553–605 (2004)

    Article  MATH  MathSciNet  Google Scholar 

  36. Passemier, D., Yao, J.-F., On estimation of the noise variance in high-dimensional probabilistic principal component analysis, submitted to J. R. Statist. Soc. Ser. B (2013) arXiv:1308.3890

  37. Wang, Q., Yao, J.-F.: On the sphericity test with large-dimensional observations. Electron. J. Stat. 7, 2164–2192 (2013)

    Article  MATH  MathSciNet  Google Scholar 

  38. Onatski, A., Moreira, M.J., Hallin, M.: Asymptotic power of sphericity tests for high-dimensional data. Ann. Stat. 41(3), 1204–1231 (2013)

    Article  MATH  MathSciNet  Google Scholar 

  39. van der Vaart, A.W.: Asymptotic Statistics, Cambridge Series in Statistical and Probabilistic Mathematics. Cambridge University Press, Cambridge (1998)

    Google Scholar 

  40. Dyson, F.J.: Statistical theory of energy levels of complex systems I-III. J. Math. Phys. 3(1), 140–175 (1962)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  41. Chen, Y., Manning, S.M.: Asymptotic level spacing of the Laguerre ensemble: a Coulomb fluid approach. J. Phys. A 27(11), 3615–3620 (1994)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  42. Chen, Y., Manning, S.M.: Distribution of linear statistics in random matrix models (metallic conductance fluctuations). J. Phys. 6(16), 3039–3044 (1994)

    MathSciNet  Google Scholar 

  43. Chen, Y., Ismail, M.E.: Thermodynamic relations of the Hermitian matrix ensembles. J. Phys. A 30(19), 6633–6654 (1997)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  44. Basor, E., Chen, Y.: Perturbed Hankel determinants. J. Phys. A 38(47), 10101–10106 (2005)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  45. Simon, S.H., Moustakas, A.L.: Crossover from conserving to lossy transport in circular random-matrix ensembles. Phys. Rev. Lett. 96(13), 136805 (2006)

    Article  ADS  Google Scholar 

  46. Vivo, P., Majumdar, S.N., Bohigas, O.: Large deviations of the maximum eigenvalue in Wishart random matrices. J. Phys. A 40(16), 4317–4337 (2007)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  47. Vivo, P., Majumdar, S.N., Bohigas, O.: Distributions of conductance and shot noise and associated phase transitions. Phys. Rev. Lett. 101(21), 216809 (2008)

    Article  ADS  Google Scholar 

  48. Dean, D.S., Majumdar, S.N.: Extreme value statistics of eigenvalues of Gaussian random matrices. Phys. Rev. E 77(4), 041108 (2008)

    Article  MathSciNet  ADS  Google Scholar 

  49. Katzav, E., Castillo, I.P.: Large deviations of the smallest eigenvalue of the Wishart-Laguerre ensemble. Phys. Rev. E 82(4), 040104 (2010)

    Article  MathSciNet  ADS  Google Scholar 

  50. Kazakopoulos, P., Mertikopoulos, P., Moustakas, A.L., Caire, G.: Living at the edge: a large deviations approach to the outage MIMO capacity. IEEE Trans. Inform. Theory 57(4), 1984–2007 (2011)

    Article  MathSciNet  Google Scholar 

  51. Li, S., McKay, M.R., Chen, Y.: On the distribution of MIMO mutual information: An in-depth Painlevé based characterization. IEEE Trans. Inform. Theory 59(9), 5271–5296 (2013)

    Article  MathSciNet  Google Scholar 

  52. Chen, Y., Haq, N.S., McKay, M.: Random matrix models, double-time Painlevé equations, and wireless relaying. J. Math. Phys. 54(6), 063506 (2013)

    Article  MathSciNet  ADS  Google Scholar 

  53. Wang, D.: The largest eigenvalue of real symmetric, Hermitian and Hermitian self-dual random matrix models with rank one external source, part I. J. Stat. Phys. 146(4), 719–761 (2012)

    Article  MATH  MathSciNet  Google Scholar 

  54. Mo, M.Y.: Rank 1 real Wishart spiked model. Commun. Pure Appl. Math. 65(11), 1528–1638 (2012)

    Article  MATH  Google Scholar 

  55. Dharmawansa, P.: Three problems related to the eigenvalues of complex non-central Wishart matrices with rank-1 mean, submitted to SIAM J. Matrix Anal. Appl. (2013) arXiv:1306.6566

  56. James, A.T.: Distributions of matrix variates and latent roots derived from normal samples. Ann. Math. Stat. 35(2), 475–501 (1964)

    Article  MATH  Google Scholar 

  57. Mehta, M.L.: Random Matrices. Pure and Applied Mathematics, 3rd edn. Elsevier/Academic Press, Amsterdam (2004)

    MATH  Google Scholar 

  58. Mauchly, J.W.: Significance test for sphericity of a normal \(n\)-variate distribution. Ann. Math. Stat. 11(2), 204–209 (1940)

    Article  MathSciNet  Google Scholar 

  59. Anderson, T.W.: An Introduction to Multivariate Statistical Analysis. Wiley Series in Probability and Statistics, 3rd edn. Wiley, Hoboken (2003)

    MATH  Google Scholar 

  60. Onatski, A., Moreira, M.J., Hallin, M.: Signal detection in high dimension: the multispiked case. Ann. Stat. 42(1), 225–254 (2014)

    Article  MATH  MathSciNet  Google Scholar 

  61. Gradshteyn, I., Ryzhik, I.: Table of Integrals, Series, and Products, Alan Jeffrey. Academic Press, Boston (2000)

    Google Scholar 

  62. Foschini, G.J., Gans, M.J.: On limits of wireless communications in a fading environment when using multiple antennas. Wirel. Pers. Commun. 6(3), 311–335 (1998)

    Article  Google Scholar 

  63. Telatar, I.E.: Capacity of multi-antenna Gaussian channels. Eur. Trans. Commun. 10(6), 585–595 (1999)

    Google Scholar 

  64. Chiani, M., Win, M.Z., Zanella, A.: On the capacity of spatially correlated MIMO Rayleigh-fading channels. IEEE Trans. Inform. Theory 49(10), 2363–2371 (2003)

    Article  MATH  MathSciNet  Google Scholar 

  65. Smith, P.J., Roy, S., Shafi, M.: Capacity of MIMO systems with semicorrelated flat fading. IEEE Trans. Inform. Theory 49(10), 2781–2788 (2003)

    Article  MATH  MathSciNet  Google Scholar 

  66. Kammoun, A., Kharouf, M., Hachem, W., Najim, J., El Kharroubi, A.: On the fluctuations of the mutual information for non centered MIMO channels: the non Gaussian case. In: Signal Processing Advances in Wireless Communications (SPAWC) (2010), pp. 1–5

  67. Hachem, W., Loubaton, P., Najim, J.: Deterministic equivalents for certain functionals of large random matrices. Ann. Appl. Probab. 17(3), 875–930 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  68. Moustakas, A.L., Simon, S.H.: Random matrix theory of multi-antenna communications: the Rician channel. J. Phys. A 38(49), 10859–10872 (2005)

    Article  MATH  MathSciNet  ADS  Google Scholar 

  69. Taricco, G.: Asymptotic mutual information statistics of separately correlated Rician fading MIMO channels. IEEE Trans. Inform. Theory 54(8), 3490–3504 (2008)

    Article  MATH  MathSciNet  Google Scholar 

  70. Kang, M., Alouini, M.-S.: Capacity of MIMO Rician channels. IEEE Trans. Wirel. Commun. 5(1), 112–122 (2006)

    Article  Google Scholar 

  71. Alfano, G., Lozano, A., Tulino, A.M., Verdú, S.: Mutual information and eigenvalue distribution of MIMO Ricean channels. In: IEEE International Symposium on Information Theory and Application (ISITA), pp. 1040–1045 (2004)

  72. Jayaweera, S.K., Poor, H.V.: On the capacity of multiple-antenna systems in Rician fading. IEEE Trans. Wirel. Commun. 4(3), 1102–1111 (2005)

    Article  Google Scholar 

  73. J. Hansen, H. Bolcskei, A geometrical investigation of the rank-\(1\) Rician MIMO channel at high SNR. In: IEEE International Symposium on Information Theory (ISIT), p. 64 (2004)

  74. McKay, M.R., Collings, I.B.: General capacity bound for spatially correlated Rician MIMO channels. IEEE Trans. Inform. Theory 51(9), 3121–3145 (2005)

    Article  MATH  MathSciNet  Google Scholar 

  75. Cui, X.W., Zhang, Q.T., Feng, Z.M.: Generic procedure for tightly bounding the capacity of MIMO correlated Rician fading channels. IEEE Trans. Commun. 53(4), 890–898 (2005)

    Article  Google Scholar 

  76. McKay, M.R., Collings, I.B.: Improved lower general bound for spatially-correlated Rician MIMO capacity. IEEE Commun. Lett. 10(3), 162–164 (2006)

    Article  Google Scholar 

  77. McKay, M.R., Smith, P.J., Collings, I.B.: New properties of complex noncentral quadratic forms and bounds on MIMO mutual information. In: IEEE International Symposium on Information Theory (ISIT), pp. 1209–1213 (2006)

  78. Jin, S., Gao, X., You, X.: On the ergodic capacity of rank-1 Ricean-fading MIMO channels. IEEE Trans. Inform. Theory 53(2), 502–517 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  79. Moustakas, A.L., Simon, S.H., Sengupta, A.M.: Mimo capacity through correlated channels in the presence of correlated interferers and noise: a (not so) large n analysis. IEEE Trans. Inform. Theory 49(10), 2545–2561 (2003)

    Article  MATH  MathSciNet  Google Scholar 

  80. Tulino, A., Verdu, S.: Asymptotic outage capacity of multiantenna channels. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Vol. 5, pp. 825–828 (2005)

  81. Hachem, W., Khorunzhiy, O., Loubaton, P., Najim, J., Pastur, L.: A new approach for mutual information analysis of large dimensional multi-antenna channels. IEEE Trans. Inform. Theory 54(9), 3987–4004 (2008)

    Article  MathSciNet  Google Scholar 

  82. Lozano, A., Tulino, A.M., Verdú, S.: High-SNR power offset in multiantenna communication. IEEE Trans. Inform. Theory 51(12), 4134–4151 (2005)

    Article  MATH  MathSciNet  Google Scholar 

  83. Schatzoff, M.: Comparative robustness of six tests in multivariate analysis of variance. J. Am. Stat. Assoc. 61(314), 415–435 (1966)

    Article  MathSciNet  Google Scholar 

  84. Olson, C.L.: Comparative robustness of six tests in multivariate analysis of variance. J. Am. Stat. Assoc. 69(348), 894–908 (1974)

    Article  MATH  Google Scholar 

  85. Walpole, R.E., Myers, R.H., Myers, S.L., Ye, K.: Probability and Statistics for Engineers and Scientists, 8th edn. Prentice Hall, New Jersey (2003)

    Google Scholar 

  86. Takemura, A.: Zonal Polynomials, Institute of Mathematical Statistics Lecture Notes-Monograph Series, 4. Institute of Mathematical Statistics, Hayward (1984)

    Google Scholar 

  87. Dyson, F.J.: An Ising ferromagnet with discontinuous long-range order. Commun. Math. Phys. 21(4), 269–283 (1971)

    Article  MathSciNet  ADS  Google Scholar 

  88. Marčenko, V.A., Pastur, L.A.: Distribution of eigenvalues in certain sets of random matrices. Mat. Sb. (N.S.) 72(114), 507–536 (1967)

    MathSciNet  Google Scholar 

  89. Olver, F.W.: Asymptotics and Special Functions, 4th edn. A K Peters, Natick (1997)

    MATH  Google Scholar 

  90. Bleistein, N., Handelsman, R.A.: Asymptotic Expansions of Integrals, 2nd edn. Dover Publications Inc., New York (1986)

    Google Scholar 

  91. Bai, Z.D.: Convergence rate of expected spectral distributions of large random matrices. II. Sample covariance matrices. Ann. Probab. 21(2), 649–672 (1993)

    Article  MATH  MathSciNet  Google Scholar 

  92. Abramowitz, M., Stegun, I.A.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, National Bureau of Standards Applied Mathematics Series. U.S. Government Printing Office, Washington DC (1964)

    Google Scholar 

  93. Tweddle, I.: James Stirling’s Methodus differentialis, Sources and Studies in the History of Mathematics and Physical Sciences, Springer-Verlag. London Ltd., London (2003)

    Google Scholar 

  94. Slater, L.J.: Confluent Hypergeometric Functions. Cambridge University Press, New York (1960)

    MATH  Google Scholar 

Download references

Acknowledgments

Thanks to Prof. Iain Johnstone at Stanford University for useful discussions in relation to the non-central multivariate \(F\) matrices (Model C) and for pointing out associated applications. The work of D. Passemier and M. R. McKay was supported by the Hong Kong Research Grants Council (RGC) under Grant number 616911. The work of Y. Chen was supported by the Macau Science Foundation Grant, under Grant number FDCT 077/2012/A3.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Damien Passemier.

Appendices

Appendix 1: Statement and Proof of Lemma 2

Lemma 2

Let \(u>0\) and \(v>0\) such that \(nu+1>0\) and \(n(v-u)\notin \mathbb {N}\). Assume that \(\mathfrak {R}z>1\) and \(\gamma \ge 0\). As \(n\rightarrow \infty \), we have

$$\begin{aligned}&\,_1F_1(nu+1,nv+1,n\gamma z) \nonumber \\& \approx \frac{1}{\sqrt{2\pi n}}\frac{\Gamma (n(u-v))\Gamma (nv+1)}{\Gamma (nu+1)}\frac{\mathrm {e}^{n\gamma z t(z)}t(z)^{nu+1}(t(z)-1)^{n(u-v)}}{\sqrt{-v(t(z)-1)^2+(u-v)(2t(z)-1)}}, \end{aligned}$$
(69)

where \(t(z)= \frac{\gamma z -v+\sqrt{(v-\gamma z)^2+4\gamma z u}}{2\gamma z}\).

Proof

Under the prescribed conditions on \(u\) and \(v\), we may use the following integral representation [94]

$$\begin{aligned} \,_1F_1(nu+1,nv+1,n\gamma z)=\frac{\Gamma (n(u-v))\Gamma (nv+1)}{2\imath \pi \Gamma (nu+1)}\oint _C\frac{t^{nu}e^{n\gamma z t}}{(t-1)^{n(u-v)+1}}\,\mathrm {d}t, \end{aligned}$$
(70)

where the contour \(C\) starts at 0, traverses anti-clockwise around 1 and returns to 0. For large \(n\), the Laplace approximation yields

$$\begin{aligned} \oint _C\frac{t^{nu}e^{n\gamma z t}}{(t-1)^{n(u-v)+1}}\,\mathrm {d}t\approx \sqrt{\frac{2\pi }{n}}\frac{q(t_0)\mathrm {e}^{-np(t_0)}}{\sqrt{p''(t_0)}}, \end{aligned}$$
(71)

for which

$$\begin{aligned} p(t)&=-\left( \gamma z t + u\text {Log}(t)+(v-u)\text {Log}(t-1)\right) \nonumber \\ q(t)&=(t-1)^{-1} \end{aligned}$$
(72)

and where \(t_0\) is the saddlepoint, which is the solution to

$$\begin{aligned} p'(t_0) = 0 . \end{aligned}$$

Thus, \(t_0\) must satisfy

$$\begin{aligned} \gamma z t_0^2+t_0(v-\gamma z)-u=0, \end{aligned}$$

with \(t_0 \notin \{0,1\)}. There is one solution which lies outside the contour for \(\mathfrak {R}z > 1\):

$$\begin{aligned} t_0(z)=\frac{\gamma z-v+\sqrt{(v-\gamma z)^2+4u\gamma z}}{2\gamma z}. \end{aligned}$$

Furthermore, we have

$$\begin{aligned} p''(t_0(z))=-\frac{-v(t_0(z)-1)^2+(u-v)(2t_0(z)-1)}{t_0(z)^2(t_0(z)-1)^2} \end{aligned}$$

so that, taking the root with the correct phase factor (see [89, Chap. 4] or [90, Chap. 7] for more details), we get

$$\begin{aligned} \frac{1}{\sqrt{p''(t_0(z))}}=\imath \frac{t_0(z)(t_0(z)-1)}{\sqrt{-v(t_0(z)-1)^2+(u-v)(2t_0(z)-1)}} . \end{aligned}$$

Substituting this quantity with (72) into (71) together with (70), we find the desired result (69). \(\square \)

Appendix 2: Equivalence Between \(\int _a^b f(x) \rho _2 (x, z)\mathrm {d}x\) and \(\int _a^b \text {Log}(z-x) \rho _1(x) \mathrm {d}x\) in (39)

We want to show that, for \(z \notin [a,b]\),

$$\begin{aligned} \int _a^b f(x) \rho _2 (x, z)\mathrm {d}x = \int _a^b \text {Log}(z-x) \rho _1(x) \mathrm {d}x, \end{aligned}$$

where

$$\begin{aligned} \int _a^b f(x) \rho _2 (x, z)\mathrm {d}x&=\frac{1}{2\pi } \int _a^b \frac{f(x)}{\sqrt{(b-x)(x-a)}}\left( \frac{\sqrt{(z-a)(z-b)}}{z-x}-1 \right) \, \mathrm {d}x \end{aligned}$$
(73)

and

$$\begin{aligned} \int _a^b \text {Log}(z-x) \rho _1(x)\, \mathrm {d}x=\frac{1}{2\pi ^2} \int _a^b \frac{\text {Log}(z-x)}{\sqrt{(b-x)(x-a)}} \left[ \mathcal {P} \int _a^b \frac{f'(y)\sqrt{(b-y)(y-a)}}{y-x} \,\mathrm {d}y \right] \, \mathrm {d}x . \end{aligned}$$
(74)

This identity is not straightforward and appears difficult to show directly using the above expressions. Thus, here we adopt an approach based on first showing that the derivative with respect to \(z\) of (73) and (74) are equal. First considering (73), we have

$$\begin{aligned} \frac{\mathrm {d}}{\mathrm {d}z}\left[ \int _a^b f(x) \rho _2 (x, z)\,\mathrm {d}x\right]&=\frac{1}{2\pi \sqrt{(z-a)(z-b)}} \int _a^b \frac{f(x)\left( (a+b)(x+z)-2ab \right) }{2(z-x)^2\sqrt{(b-x)(x-a)}} \, \mathrm {d}x. \end{aligned}$$
(75)

Now taking the derivative of (74), we get

$$\begin{aligned}&\frac{\mathrm {d}}{\mathrm {d}z}\left[ \int _a^b \text {Log}(z-x) \rho _1(x)\,\mathrm {d}x \right] \\& = \int _a^b \frac{\sqrt{(b-y)(y-a)}}{2\pi ^2}f'(y) \left[ \mathcal {P} \int _a^b \frac{\mathrm {d}x}{(x-y)(z-x)\sqrt{(b-x)(x-a)}} \right] \, \mathrm {d}y \\& =\int _a^b \frac{\sqrt{(b-y)(y-a)}}{2\pi ^2(z-y)}f'(y) \left[ \mathcal {P} \int _a^b \frac{(z-x)^{-1}+(x-y)^{-1}}{\sqrt{(b-x)(x-a)}} \mathrm {d}x\right] \, \mathrm {d}y, \end{aligned}$$

which, after applying the identities (86) and (89), yields

$$\begin{aligned} \frac{\mathrm {d}}{\mathrm {d}z}\left[ \int _a^b \text {Log}(z-x) \rho _1(x)\,\mathrm {d}x \right]&=\frac{1}{2\pi \sqrt{(z-a)(z-b)}}\int _a^b \frac{\sqrt{(b-y)(y-a)}}{y-z}f'(y)\,\mathrm {d}y. \end{aligned}$$

Application of integration by parts to the last integral gives

$$\begin{aligned} \frac{\mathrm {d}}{\mathrm {d}z}\left[ \int _a^b \text {Log}(z-x) \rho _1(x)\, \mathrm {d}x \right]&=\frac{1}{2\pi \sqrt{(z-a)(z-b)}}\ \int _a^b \frac{f(y)\left( (a+b)(y+z)-2ab \right) }{2(z-y)^2\sqrt{(b-y)(y-a)}} \, \mathrm {d}y, \end{aligned}$$

which is the same as (75). So we have proved that

$$\begin{aligned} \int _a^b \text {Log}(z-x) \rho _1(x)\, \mathrm {d}x = \int _a^b f(x) \rho _2 (x, z)\mathrm {d}x + \text {Constant} . \end{aligned}$$
(76)

Now, note that with \(z\) such that \(\mathfrak {I}z=0\), as \(\mathfrak {R}z \rightarrow \infty \),

$$\begin{aligned} \frac{\sqrt{(z-a)(z-b)}}{z-x}-1 = \frac{x-\frac{a+b}{2}}{z}+ O\left( \frac{1}{z^2} \right) . \end{aligned}$$

Plugging this expression into (73) gives

$$\begin{aligned} \int _a^b f(x) \rho _2 (x, z)\mathrm {d}x = \frac{1}{z} \int _a^b \frac{f(x)}{\sqrt{(b-x)(x-a)}}\left( x-\frac{a+b}{2} \right) \, \mathrm {d}x + O\left( \frac{1}{z^2} \right) \end{aligned}$$

which tends to zero as \(\mathfrak {R}z \rightarrow \infty \). Furthermore, with

$$\begin{aligned} \text {Log}(z-x) = \text {Log}(z)-\sum _{k=1}^\infty \frac{x^k}{z^k} , \end{aligned}$$

the expression in (74) becomes, upon interchanging the integrals,

$$\begin{aligned} \int _a^b \text {Log}(z-x) \rho _1(x)\, \mathrm {d}x&=\frac{1}{2\pi ^2} \int _a^b f'(y)\sqrt{(b-y)(y-a)} \left[ \mathcal {P} \int _a^b \frac{\text {Log}(z)\,\mathrm {d}x}{(y-x)\sqrt{(b-x)(x-a)}}\right. \\& -\left. \sum _{k=1}^\infty \frac{\mathcal {P}}{z^k}\int _a^b\frac{x^k\,\mathrm {d}x}{(y-x)\sqrt{(b-x)(x-a)}} \right] \, \mathrm {d}y . \end{aligned}$$

Here, the first principal value integral is zero by (86), whilst the remaining terms tends to zero when \(z\) is such that \(\mathfrak {R}z \rightarrow \infty \) and \(\mathfrak {I}z=0\).

Consequently, taking \(z\) such that \(\mathfrak {R}z \rightarrow \infty \) and \(\mathfrak {I}z=0\) in (76), we find that the constant term is zero, thus proving the result.

Appendix 3: Useful Formulas

For the derivations of our results, we will require numerous integrals; these are summarized in (7793). Note that for all definite integrals involving the variable \(t\), these are valid for \(\mathfrak {R}t >b\), while in all cases we assume \(0<a<b\).

$$\begin{aligned}&\int _a^b \frac{\text {Log}(x+t)}{\sqrt{(b-x)(x-a)}}\, \mathrm {d}x=2\pi \text {Log}\left( \frac{\sqrt{t+a}+\sqrt{t+b}}{2} \right) \end{aligned}$$
(77)
$$\begin{aligned}&\int _a^b \frac{\mathrm {d}x}{(x+t)\sqrt{(b-x)(x-a)}} =\frac{\pi }{\sqrt{(t+a)(t+b)}} \end{aligned}$$
(78)
$$\begin{aligned}&\int _a^b \frac{\mathrm {d}x}{\sqrt{(b-x)(x-a)}}=\pi \end{aligned}$$
(79)
$$\begin{aligned}&\int _a^b \frac{x\,\mathrm {d}x}{\sqrt{(b-x)(x-a)}}=\pi \frac{a+b}{2} \end{aligned}$$
(80)
$$\begin{aligned}&\int _a^b \frac{\text {Log}(x+t)}{\sqrt{(b-x)(x-a)}x}\,\mathrm {d}x=\frac{\pi }{\sqrt{ab}}\text {Log}\left( \frac{(\sqrt{ab}+\sqrt{(t+a)(t+b)})^2-t^2}{(\sqrt{a}+\sqrt{b})^2}\right) \end{aligned}$$
(81)
$$\begin{aligned}&\int _a^b \frac{x\text {Log}(x+t)}{\sqrt{(b-x)(x-a)}}\,\mathrm {d}x=\pi \frac{(\sqrt{a+t}-\sqrt{b+t})^2}{2}\nonumber \\&\qquad \quad +\pi \frac{a+b}{2}\text {Log}\left( \frac{2(t+(a+t)(b+t))+a+b}{4}\right) \end{aligned}$$
(82)
$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)}{\sqrt{(b-x)(x-a)}}\, \mathrm {d}x=2\pi \text {Log}\left( \frac{\sqrt{1-a}+\sqrt{1-b}}{2} \right) \end{aligned}$$
(83)
$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)}{\sqrt{(b-x)(x-a)}x}\,\mathrm {d}x=\frac{\pi }{\sqrt{ab}}\text {Log}\left( \frac{1-(\sqrt{ab}-\sqrt{(1-a)(1-b)})^2}{(\sqrt{a}+\sqrt{b})^2}\right) \end{aligned}$$
(84)
$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)}{\sqrt{(b-x)(x-a)}(x-1)}\,\mathrm {d}x=\frac{2\pi }{\sqrt{(1-a)(1-b)}}\text {Log}\left( \frac{1}{2\sqrt{1-a}}+\frac{1}{2\sqrt{1-b}}\right) \end{aligned}$$
(85)
$$\begin{aligned}&\mathcal {P}\int _a^b \frac{1}{(y-x)\sqrt{(b-x)(x-a)}}\,\mathrm {d}x=0\end{aligned}$$
(86)
$$\begin{aligned}&\mathcal {P} \int _a^b \frac{\sqrt{(b-x)(x-a)}}{x(y-x)} \, \mathrm {d}x =\pi \left( 1-\frac{\sqrt{ab}}{y} \right) \end{aligned}$$
(87)
$$\begin{aligned}&\mathcal {P} \int _a^b \frac{\sqrt{(b-y)(y-a)}}{(1-y)(x-y)} \,\mathrm {d}y = \pi \left( \frac{\sqrt{(1-a)(1-b)}}{1-x}-1\right) \end{aligned}$$
(88)
$$\begin{aligned}&\int _a^b \frac{1}{(x-t)\sqrt{(b-x)(x-a)}}\,\mathrm {d}x= - \frac{\pi }{\sqrt{(t-a)(t-b)}} \end{aligned}$$
(89)
$$\begin{aligned}&\int _a^b \frac{1}{\sqrt{(b-x)(x-a)}}\left( \frac{\sqrt{(z-a)(z-b)}}{t-x}-1 \right) \, \mathrm {d}x=0\end{aligned}$$
(90)
$$\begin{aligned}&\mathcal {P} \int _a^b \frac{\sqrt{(b-x)(x-a)}}{y-x} \,\mathrm {d}x = \pi \left( y-\frac{a+b}{2} \right) \end{aligned}$$
(91)

Moreover, for \(z \in \mathbb {C}\) and \(b<1\),

$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)}{(x-z)\sqrt{(b-x)(x-a)}} \,\mathrm {d}x\nonumber \\&\qquad \quad =\frac{\pi }{A} \text {Log}\left( \frac{2A+2z-a-b}{2A\sqrt{(1-a)(1-b)}+z(2-a-b)-a-b+2ab}\right) \end{aligned}$$
(92)

and for \(z \in \mathbb {C}\),

$$\begin{aligned}&\int _a^b \frac{\text {Log}(x+t) \, \mathrm {d}x}{(x-z)\sqrt{(b-x)(x-a)}}\nonumber \\& =\frac{\pi }{A}\text {Log}\left( \frac{2A+(a+b-2z)(t+z)+2A\sqrt{A^2+(a+b-2z)(t+z)+(t+z)^2}}{(t+z)^2(a+b-2z+2A)}\right) \end{aligned}$$
(93)

with \(A=\sqrt{(z-a)(z-b)}\).

Equations (7788) are given in [18], whilst (89) is a slight modification of (78), and (90) follows using (78) and (79). The expression (91) follows upon multiplying the numerator and the denominator of the integrand by \(\sqrt{(b-x)(x-a)}\), applying a partial fraction decomposition, then invoking (79), (80) and (86). For (92) and (93), the derivations are more involved, and we describe each in turn.

For (92), we use the parametrization

$$\begin{aligned} \text {Log}(1-x) = \int _0^1 \frac{x}{\lambda x-1} \,\mathrm {d}\lambda \end{aligned}$$

and the partial fraction decomposition

$$\begin{aligned} \frac{x}{(x-1/\lambda )(x-z)}=\frac{1/\lambda }{(x-1/\lambda )(1/\lambda -z)}-\frac{z}{(x-z)(1/\lambda -z)} \end{aligned}$$

to arrive at

$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)\,\mathrm {d}x}{(x-z)\sqrt{(b-x)(x-a)}} =\int _0^1 \frac{\mathrm {d}\lambda }{1-\lambda z} \int _a^b \left( \frac{1/\lambda }{x-1/\lambda } + \frac{z}{z-x} \right) \frac{\mathrm {d}x}{\sqrt{(b-x)(x-a)}}\\&\quad =\, \pi \int _0^1 \left( \frac{-1}{\sqrt{(1-\lambda a)(1-\lambda b)}} + \frac{z}{\sqrt{(z-a)(z-b)}}\right) \frac{\mathrm {d}\lambda }{ 1-\lambda z}. \end{aligned}$$

The last equation was obtained by invoking (78) and (89). From a further change of variable \(x=1- \lambda z\), we have

$$\begin{aligned}&\int _a^b \frac{\text {Log}(1-x)}{(x-z)\sqrt{(b-x)(x-a)}} \,\mathrm {d}x = \pi \int _{1-z}^1 \left( \frac{-1}{\sqrt{(z-a+ax)(z-b+bx)}} + \frac{1}{A}\right) \frac{\mathrm {d}x}{x}\\&\quad =\frac{\pi }{A} \Bigg ( -\text {Log}(1-z) \\&\qquad + \left. \left[ \text {Log}\left( \frac{2A\sqrt{(z-a+ax)(z-b+bx)}+2A^2+x(z(a+b)-2ab)}{x}\right) \right] _{1-z}^1\right) \\&\quad =\frac{\pi }{A} \text {Log}\left( \frac{2A+2z-a-b}{2A\sqrt{(1-a)(1-b)}+z(2-a-b)-a-b+2ab}\right) . \end{aligned}$$

Now consider (93). In this case, we use the relation

$$\begin{aligned} \text {Log}(x+t) = \text {Log}t + \int _0^1 \frac{x}{t+ yx} \mathrm {d}y \end{aligned}$$

to give

$$\begin{aligned}&\int _a^b \frac{\text {Log}(x+t) \, \mathrm {d}x}{(x-z)\sqrt{(b-x)(x-a)}}\\& =\int _a^b \frac{\text {Log}t \, \mathrm {d}x}{(x-z)\sqrt{(b-x)(x-a)}}+\int _0^1 \left[ \int _a^b \frac{1}{\sqrt{(b-x)(x-a)}}\frac{x}{t+yx}\frac{\mathrm {d}x}{x-z}\right] \mathrm {d}y{.} \end{aligned}$$

The first integral is given by (89), whereas the double integral is

$$\begin{aligned}&\int _0^1 \left[ \int _a^b \frac{1}{\sqrt{(b-x)(x-a)}} \frac{x}{x+\frac{t}{y}}\frac{\mathrm {d}x}{x-z} \right] \frac{\mathrm {d}y}{y} \\& = \int _0^1\left[ \int _a^b \frac{1}{\sqrt{(b-x)(x-a)}} \left( \frac{z}{x-z}+ \frac{\frac{t}{y}}{x+\frac{t}{y}}\right) \mathrm {d}x \right] \frac{\mathrm {d}y}{y \left( \frac{t}{y}+z\right) }\\& = \int _0^1 \frac{z}{y\left( \frac{t}{y}+z\right) }\left[ \int _a^b \frac{ \, \mathrm {d}x}{(x-z)\sqrt{(b-x)(x-a)}}\right] \mathrm {d}y \\&\quad +\int _0^1 \frac{\frac{t}{y}}{y \left( \frac{t}{y}+z\right) } \left[ \int _a^b \frac{1}{\sqrt{(b-x)(x-a)}} \frac{\mathrm {d}x}{x+\frac{t}{y}} \right] \mathrm {d}y . \end{aligned}$$

The first term on the right-hand side above is obtained using (89), whereas the second is

$$\begin{aligned}&\int _0^1 \frac{1}{y}\left[ \frac{\frac{t}{y}}{\frac{t}{y}+z} \int _a^b \frac{1}{\sqrt{(b-x)(x-a)}} \frac{\mathrm {d}x}{x+\frac{t}{y}}\right] \mathrm {d}y\\& = \pi \int _0^1 \frac{1}{y}\frac{\frac{t}{y}}{\frac{t}{y}+z} \frac{ \mathrm {d}y}{\sqrt{\left( b+\frac{t}{y}\right) \left( a+\frac{t}{y}\right) }} \quad \quad \text {[using (78)]}\\& = \pi \int _{t+z}^{\infty } \frac{\, \mathrm {d}x}{x\sqrt{(x-z+a)(x-z+b)}} \quad \quad [\text {setting} x= t/y +z]\\& = \pi \int _{t+z}^{\infty } \frac{\, \mathrm {d}x}{x\sqrt{-(b-z)(z-a)+(b+a-2z)x+x^2}}\\& = \frac{\pi }{A} \text {Log}\left( \frac{2A+(a+b-2z)(t+z)+2A\sqrt{A^2+(a+b-2z)(t+z)+(t+z)^2}}{(t+z)(a+b-2z+2A)}\right) , \end{aligned}$$

using the first identity in [61, Eq. 2.266]. Combining the previous calculations, we get the result.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Passemier, D., McKay, M.R. & Chen, Y. Asymptotic Linear Spectral Statistics for Spiked Hermitian Random Matrices. J Stat Phys 160, 120–150 (2015). https://doi.org/10.1007/s10955-015-1233-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10955-015-1233-x

Keywords

Navigation