In capacitance-voltage (C-V) profiling of a semiconductor, the standard interpretation of the measured differential capacitance yields an apparent carrier concentration n^(x)\hat{n}(x) that is a Debye-averaged, smoothed version of the true free-carrier profile n(x)n(x): each small increment of reverse bias removes an incremental charge distribution whose mean position is fixed exactly by the capacitance and whose width is of the order of the Debye length LDL_D, the characteristic electrostatic screening length of the free carriers. Kroemer and Chien (1981) established exact constraints for uniformly doped, non-degenerate material — conservation of charge and of the first moment between n^\hat{n} and nn over any interval bounded by uniform regions, and a centered second moment of the averaging kernel equal to 2LD22L_D^{2} (RMS averaging distance 2 LD\sqrt{2}\,L_D) — and asked whether such constraints can be turned into a reconstruction. They left open: (i) a numerical algorithm that approximately reconstructs n(x)n(x) from n^(x)\hat{n}(x) for non-uniform profiles, for which the uniform-background kernel is only a first-order approximation, with accuracy and convergence unexplored; (ii) reliable extraction of graded-step steepness in the regime where their moment-based formula suffers near-cancellation of comparable terms; and (iii) the general inversion itself, for which trial-and-error forward simulation was the only method. Four decades of later work has systematized trial-and-error into parametric inverse modeling, reached sub-Debye resolution only by changing the measurement (destructive electrochemical profiling with etch steps), and formalized the ill-posedness (extreme sensitivity of the recovered profile to small perturbations of the data) of related inverse doping problems driven by current data — but no general reconstruction of n(x)n(x) from a single standard C-V profile, with demonstrated accuracy and convergence, has been published. The problem remains open.

Background

C-V profiling is the standard electrical method for measuring the depth profile of the free-carrier (and, indirectly, dopant) concentration in a semiconductor. A Schottky barrier (the rectifying energy barrier formed where a metal contacts a semiconductor), p-n junction, MOS (metal-oxide-semiconductor) capacitor, or electrolyte contact is reverse-biased, and the small-signal differential capacitance per unit area C(V)C(V) is measured. In the depletion approximation — an abrupt boundary at depth ww between a fully depleted region and a neutral region — the carrier concentration is read off as

n^(w)=[qϵ2ddV(1C2)]−1,w=ϵC,\hat{n}(w)=\left[\frac{q\epsilon}{2}\frac{d}{dV}\left(\frac{1}{C^{2}}\right)\right]^{-1},\qquad w=\frac{\epsilon}{C},

where qq is the elementary charge and ϵ\epsilon the semiconductor permittivity.

The depletion approximation is not exact: free carriers are distributed over the Debye length LD=(ϵkT/q2n)1/2L_D=(\epsilon kT/q^{2}n)^{1/2} around the depletion edge (kk is Boltzmann's constant, TT the temperature, nn the local carrier concentration), so the measured n^\hat{n} is a smeared average of the true profile n(x)n(x) over roughly a Debye length — the Debye averaging analyzed by Johnson and Panousis (1971) by forward numerical simulation. For uniform, non-degenerate material (non-degenerate meaning that the carriers obey classical Maxwell-Boltzmann rather than degenerate Fermi-Dirac statistics) the exact 1/C21/C^{2}-versus-VV relation differs from the depletion-approximation one by a voltage offset of kT/qkT/q (the Gummel-Scharfetter correction; Scharfetter and Gummel 1969). Kennedy and co-workers (1968, 1969) treated the related but distinct problem of recovering the dopant distribution from the true electron profile; the Debye averaging between n(x)n(x) and n^(x)\hat{n}(x) is the more severe limitation, particularly for ion-implanted layers, delta-doped layers (ultrathin, high-concentration dopant sheets), and isotype heterojunctions (junctions between two semiconductors of the same doping type), where C-V profiling is the workhorse technique for extracting conduction-band offsets (the energy difference between the conduction-band edges of the two joined semiconductors) and interface charge.

Kroemer and Chien (1981) gave this averaging a precise structure. When the reverse bias is increased by a small increment, a distribution of electrons — the incremental displaced electron distribution (IDED) — is removed from the semiconductor. Exactly: (i) the mean position of the IDED is ⟨x⟩=ϵ/C\langle x\rangle=\epsilon/C, so a capacitance measurement locates the mean of the incremental charge; and (ii) n^\hat{n} is constructed as the sequence of rectangular strips that conserve each IDED's charge and first moment. Two identities then hold for any interval (a,b)(a,b) that spans a non-uniform region and is bounded by regions where nn is uniform (so that n^=n\hat{n}=n there):

∫abn^(x) dx=∫abn(x) dx,∫abn^(x) x dx=∫abn(x) x dx,\int_a^b \hat{n}(x)\,dx=\int_a^b n(x)\,dx,\qquad \int_a^b \hat{n}(x)\,x\,dx=\int_a^b n(x)\,x\,dx,

i.e. charge and first moment are conserved between the apparent and true profiles. For uniformly doped non-degenerate material the centered second moment of the IDED per unit displaced charge is exactly 2LD22L_D^{2}, so the RMS averaging distance is 2 LD\sqrt{2}\,L_D; for non-uniform profiles this becomes a first-order local approximation, with the Debye length evaluated at the mean position of the IDED. Combining moment conservation with two expressions for the built-in voltage of a graded (error-function) doping step,

nd(x)=nˉ−12 δn erf(x2L),n_d(x)=\bar{n}-\tfrac{1}{2}\,\delta n\,\mathrm{erf}\left(\frac{x}{2L}\right),

Kroemer and Chien derived a formula that extracts the gradient steepness LL (with nˉ\bar{n} the mean donor concentration and δn\delta n the concentration change across the step) directly from the measured n^\hat{n} without full reconstruction — but for very steep gradients the two terms on the right-hand side nearly cancel, and their difference may fall below the experimental error.

The inverse problem — deducing n(x)n(x) from a measured n^(x)\hat{n}(x) — cannot be solved by reading these formulas backward: the averaging kernel itself depends on the unknown profile through the local Debye length and the self-consistent depletion-region electrostatics, so the inversion is nonlinear. In 1981 the only general procedure was trial-and-error: guess n(x)n(x), solve Poisson's equation forward, compare simulated and measured n^\hat{n}, and iterate, with uncertain convergence. The open question is whether the exact constraints (moment conservation, the quantitative averaging width) and modern inverse-problem methodology can be turned into a general, reliable reconstruction of n(x)n(x) from a single standard C-V profile.

Problem Statement

Starting from a measured apparent carrier profile n^(x)\hat{n}(x) — the Debye-averaged, smoothed version of the true free-carrier profile n(x)n(x) that the standard interpretation of the measured differential capacitance yields — obtained by a standard, non-destructive single-contact C-V (capacitance-voltage) measurement on a semiconductor structure, with the forward map n↦n^n\mapsto\hat{n} given by the electrostatics of the reverse-biased junction (Poisson's equation with the material's carrier statistics — Maxwell-Boltzmann or Fermi-Dirac — for the classical profile classes, and a self-consistent Schrödinger-Poisson description for quantum-confined layers), the problem is the general inversion of Debye averaging: determine whether the exact constraints and modern inverse-problem methodology can be turned into a reconstruction method — a numerical algorithm or analytical procedure — that recovers the true carrier profile n(x)n(x) from n^(x)\hat{n}(x) for non-uniform profiles, and, if so, construct one together with a quantitative analysis of its accuracy and convergence, or else characterize the obstruction that bars such a general reconstruction.

The known exact ingredients are the charge- and first-moment-conservation identities (on any depth interval whose ends lie in uniformly doped regions, the integral of n^\hat{n} equals that of nn, and the integral of x n^(x)x\,\hat{n}(x) equals that of x n(x)x\,n(x)), and the uniform-background second-moment width (RMS averaging distance 2 LD\sqrt{2}\,L_D, where LDL_D is the local Debye length — the characteristic electrostatic screening length of the free carriers), whose extension to non-uniform profiles is only a first-order local approximation. The reconstruction is sought for the profile classes, carrier statistics (non-degenerate, meaning classical Maxwell-Boltzmann statistics, or degenerate Fermi-Dirac statistics), and structures it covers — uniformly doped, graded, delta-doped (ultrathin dopant sheets), isotype heterojunctions carrying interface charge (junctions between two semiconductors of the same doping type), and quantum-confined layers (layers so thin that the carriers occupy discrete quantized energy levels, so their spatial distribution is set by the quantum-mechanical wave function rather than by classical thermal statistics) — over a physically justified subset of these classes, with the reconstruction's accuracy and convergence analyzed over that subset.

References

  1. Herbert Kroemer and Wu-Yi Chien (1981). On the theory of Debye averaging in the C-V profiling of semiconductors. Solid-State Electronics 24(7), 655-660. DOI: 10.1016/0038-1101(81)90195-7. https://doi.org/10.1016/0038-1101(81)90195-7
  2. W. C. Johnson and P. T. Panousis (1971). The influence of Debye length on the C-V measurement of doping profiles. IEEE Transactions on Electron Devices ED-18(10), 965-973. DOI: 10.1109/T-ED.1971.17311. https://doi.org/10.1109/T-ED.1971.17311
  3. David P. Kennedy, P. C. Murley and W. Kleinfelder (1968). On the Measurement of Impurity Atom Distributions in Silicon by the Differential Capacitance Technique. IBM Journal of Research and Development 12(5), 399-409. DOI: 10.1147/rd.125.0399. https://doi.org/10.1147/rd.125.0399
  4. David P. Kennedy and R. R. O'Brien (1969). On the Measurement of Impurity Atom Distributions by the Differential Capacitance Technique (Letter to the Editor). IBM Journal of Research and Development 13(2), 212-214. DOI: 10.1147/rd.132.0212. https://doi.org/10.1147/rd.132.0212
  5. D. L. Scharfetter and H. K. Gummel (1969). Large-signal analysis of a silicon Read diode oscillator. IEEE Transactions on Electron Devices 16(1), 64-77. DOI: 10.1109/T-ED.1969.16566. https://doi.org/10.1109/T-ED.1969.16566
  6. Herbert Kroemer (1985). Determination of heterojunction band offsets by capacitance-voltage profiling through nonabrupt isotype heterojunctions. Applied Physics Letters 46(5), 504-505. DOI: 10.1063/1.95572. https://doi.org/10.1063/1.95572
  7. E. F. Schubert, R. F. Kopf, J. M. Kuo, H. S. Luftman and P. A. Garbinski (1990). Spatial resolution of the capacitance-voltage profiling technique on semiconductors with quantum confinement. Applied Physics Letters 57(5), 497-499. DOI: 10.1063/1.103632. https://doi.org/10.1063/1.103632
  8. Y. K. Yeo, G. H. Gainer, Jong Hyun Kim and R. L. Hengehold (1990). Evaluation of carrier distributions in a depletion region of semiconductor measured by a capacitance-voltage profile method. Applied Physics Letters 56(1), 75-77. DOI: 10.1063/1.102656. https://doi.org/10.1063/1.102656
  9. T. Toyabe, H. Matsuo, S. Yamamoto and H. Masuda (1989). Determination of doping profile in sub-Debye-length region. International Electron Devices Meeting (IEDM) Technical Digest, 699-702. DOI: 10.1109/IEDM.1989.74375. https://doi.org/10.1109/IEDM.1989.74375
  10. X. H. Fu and J. N. Chen (1992). A point-by-point multiple-sweep numerical algorithm for dopant profiling based on C-V data. Solid-State Electronics 35(2), 181-185. DOI: 10.1016/0038-1101(92)90058-K. https://doi.org/10.1016/0038-1101(92)90058-K
  11. M. Ogawa, H. Matsubayashi, H. Ohta and T. Miyoshi (1995). Accurate parameter extraction of heterojunctions based on inverse C-V simulation. Solid-State Electronics 38(6), 1197-1207. DOI: 10.1016/0038-1101(94)00240-G. https://doi.org/10.1016/0038-1101(94)00240-G
  12. Cheol Moon, Byung-Doo Choe, S. D. Kwon and Hanjin Lim (1998a). Spatial resolution of capacitance-voltage profiles in quantum well structures. Applied Physics Letters 72(10), 1196-1198. DOI: 10.1063/1.121011. https://doi.org/10.1063/1.121011
  13. Cheol Moon, Byung-Doo Choe, S. D. Kwon, H. K. Shin and Hanjin Lim (1998b). Electron distribution and capacitance-voltage characteristics of n-doped quantum wells. Journal of Applied Physics 84(5), 2673-2683. DOI: 10.1063/1.368380. https://doi.org/10.1063/1.368380
  14. M. F. Kokorev, N. Maleev and D. V. Pakhnin (2000). Inverse Modeling for C-V Profiling of Modulated-Doped Semiconductor Structures. In Technical Proceedings of the 2000 International Conference on Modeling and Simulation of Microsystems (MSM 2000), San Diego, pp. 392-395, NSTI/TechConnect Briefs. No DOI registered. https://briefs.techconnect.org/wp-content/volumes/MSM2000/pdf/T54.06.pdf
  15. V. I. Shashkin, I. R. Karetnikova, A. V. Murel, I. S. Nefedov and I. A. Shereshevskii (2000). Approach to electrochemical C-V profiling in semiconductor with sub-Debye-length resolution. IEEE Transactions on Electron Devices 47(6), 1221-1224. DOI: 10.1109/16.842965. https://doi.org/10.1109/16.842965
  16. David A. Gore and Donald A. Drew (1994). The one-dimensional inverse doping problem in semiconductor device modeling. Inverse Problems in Engineering 1(1), 27-43. DOI: 10.1080/174159794088027571. https://doi.org/10.1080/174159794088027571
  17. Yingda Cheng, Irene M. Gamba and Kui Ren (2011). Recovering doping profiles in semiconductor devices with the Boltzmann-Poisson model. Journal of Computational Physics 230(9), 3391-3412. DOI: 10.1016/j.jcp.2011.01.034. https://doi.org/10.1016/j.jcp.2011.01.034
  18. Christophe A. Hurni, Herbert Kroemer, Umesh K. Mishra and James S. Speck (2012). Capacitance-voltage profiling on polar III-nitride heterostructures. Journal of Applied Physics 112(8), 083704. DOI: 10.1063/1.4757940. https://doi.org/10.1063/1.4757940
  19. Azzouz Sellai (2012). Electrical characteristics of InAs self-assembled quantum dots embedded in GaAs using admittance spectroscopy. Journal of Nanophotonics 6(1), 063502. DOI: 10.1117/1.JNP.6.063502. https://doi.org/10.1117/1.JNP.6.063502
  20. A. F. Yaremchuk (2005). A contribution to the theory of the C-V technique for the evaluation of carrier-concentration profiles in semiconductors. Applied Physics A 80, 881-883. DOI: 10.1007/s00339-003-2337-8. https://doi.org/10.1007/s00339-003-2337-8
  21. Ya. V. Ivanova, G. E. Yakovlev, V. I. Zubkov and A. Solomnikova (2019). Emission processes of the interaction between the quantum well and donor delta layer in heterostructures for pHEMTs. Turkish Journal of Physics 43(3), 243-251. DOI: 10.3906/FIZ-1812-12. https://doi.org/10.3906/FIZ-1812-12
  22. Mathias Nyman, Oskar J. Sandberg, Staffan Dahlström, Donato Spoltore, Christian Körner, Yadong Zhang, Stephen Barlow, Seth R. Marder, Karl Leo, Koen Vandewal and Ronald Österbacka (2017). Doping-induced carrier profiles in organic semiconductors determined from capacitive extraction-current transients. Scientific Reports 7, 5397. DOI: 10.1038/s41598-017-05499-3. https://doi.org/10.1038/s41598-017-05499-3
  23. G. E. Yakovlev and V. I. Zubkov (2021). Integration of electrochemical capacitance-voltage characteristics: a new procedure for obtaining free charge carrier depth distribution profiles with high resolution. Journal of Solid State Electrochemistry 25(3), 797-802 (published online November 2020). DOI: 10.1007/s10008-020-04855-0. https://doi.org/10.1007/s10008-020-04855-0
  24. S. Piani, Patricio Farrell, Wenyu Lei, Nella Rotundo and Luca Heltai (2024). Data-driven solutions of ill-posed inverse problems arising from doping reconstruction in semiconductors. Applied Mathematics in Science and Engineering 32(1), article 2323626. DOI: 10.1080/27690911.2024.2323626. https://doi.org/10.1080/27690911.2024.2323626