Pith. sign in

REVIEW 4 major objections 6 minor 15 references

3-D Atomic Mapping of Interfacial Roughness and its Spatial Correlation Length in sub-10 nm Superlattices

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Atom-probe maps of buried Si/SiGe interfaces show growth temperature sets their roughness and correlation length, and that optical models must include this broadening.

desk verdict First direct 3D APT measurement of buried Si/SiGe interface correlation lengths is likely novel and worth engaging, but the reported sub-0.3 nm roughness sits close to the method's voxel-statistics floor and the temperature trend is confounded. read the letter →

arxiv 1908.00874 v1 pith:F45UTPBN submitted 2019-08-02 cond-mat.mtrl-sci

Samik Mukherjee , Anis Attiaoui , Matthias Bauer , Oussama Moutanabbir This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords silicon-germaniumsuperlatticesatomprobetomographyinterfacialroughnessinterfacecorrelationlengthwidthspectroscopicellipsometryburiedinterfacesheight-heightfunction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to measure, directly and in three dimensions, the roughness of interfaces buried inside sub-10 nm Si/SiGe superlattices, a quantity that shapes electronic and optical behavior but has usually been treated as a fitting parameter. Using atom probe tomography, the authors construct iso-concentration surfaces at each buried interface and analyze their height-height correlations. They report that the root-mean-square roughness is about 0.3 nm for growth at 650 °C and about 0.2 nm at 500 °C, with horizontal correlation lengths of about 8 nm and 10 nm respectively. They also find about 30% narrower interfacial widths at the lower growth temperature and attribute the difference to thermally activated atomic exchange at the growth surface. Finally, by comparing optical models with increasing interfacial complexity, they show that the atomic-level roughening must be included to reproduce spectroscopic ellipsometry data; models that omit the interface systematically overestimate layer thicknesses.

What carries the argument

The central object is the iso-concentration surface: a three-dimensional surface drawn at 50% of the mean Ge concentration within each SiGe layer, taken to represent each buried interface. Its height map $z(x,y)$ feeds the height-height correlation function $H(\tau)=\langle |z(x,y)-z(x',y')|^2\rangle$, whose fit to $2\sigma^2[1-\exp(-(\tau/\xi)^{2\alpha})]$ yields the RMS roughness $\sigma$ and the horizontal correlation length $\xi$. A second device is the sigmoid fit $c(x)=c_0+d_0/[1+\exp(-(x_0\pm x)/\mathcal{L})]$ to one-dimensional concentration profiles, whose length parameter $\mathcal{L}$ defines the interfacial width. The optical counterpart is a graded interfacial alloy model in which each interface is a graded layer with Si content following the same sigmoid function, discretized into 21 sublayers; this is the mechanism that connects the measured atomic-level broadening to the optical response.

What would settle it

A reconstruction simulation or a cross-check with a reconstruction-independent probe would settle the claim: start from a model superlattice with prescribed $\sigma$ and $\xi$, simulate atom probe evaporation with different evaporation fields for Si and SiGe, reconstruct the data, apply the iso-concentration-surface analysis, and see whether the input values are recovered; alternatively, measure identical superlattices with grazing-incidence small-angle X-ray scattering or scanning tunneling microscopy and compare the extracted roughness and correlation lengths with the atom probe values.

Watch

Extended reading notes

Core claim

The central claim is that atom probe tomography, applied to Si/SiGe superlattices with individual layers 1.5–7.5 nm thick, can map the vertical height of every buried interface as a three-dimensional iso-concentration surface, and that from these surfaces the RMS roughness $\sigma$ and horizontal correlation length $\xi$ can be extracted quantitatively. For interfaces defined at 50% of the mean Ge content, the height-height correlation function $H(\tau)$ follows the phenomenological form $H_{\mathrm{fit}}(\tau)=2\sigma^2[1-\exp(-(\tau/\xi)^{2\alpha})]$, yielding mean $\sigma = 0.3$ nm ($\pm11.5\%$) and $\xi = 8.1$ nm ($\pm5.8\%$) for growth at 650 °C, and $\sigma = 0.2$ nm ($\pm13\%$) and $\xi = 10.1$ nm ($\pm6.2\%$) for growth at 500 °C. The authors further claim that the one-dimensional Ge concentration profiles across the same interfaces give average interfacial widths of about 1.0 nm at 650 °C and 0.7 nm at 500 °C, with no asymmetry between Si→SiGe and SiGe→Si transitions, indicating suppressed Ge segregation. On the optical side, they show that a graded-alloy interfacial model reproduces the ellipsometric spectra better than models that omit or homogenize the interfaces, and that omitting interfacial broadening systematically overestimates layer thicknesses.

Load-bearing premise

The load-bearing assumption is that the atom probe reconstruction, after iterative calibration so that its layer thicknesses match transmission electron microscopy within 5%, preserves true vertical heights and horizontal separations at the buried interfaces to better than about 0.2 nm; if reconstruction artifacts shift heights at this scale, the reported roughness and correlation lengths would be measurement artifacts rather than physical interface properties.

Editorial extensions

If this is right

  • Interface roughness and correlation length can be obtained directly from atom probe data for buried interfaces, replacing the practice of treating the correlation length as a free fitting parameter in scattering and transport models.
  • Growth temperature controls interface morphology: lowering the growth temperature from 650 °C to 500 °C reduces the RMS roughness by about 30% and the interfacial width by about 30%, while increasing the horizontal correlation length by about 24%.
  • The absence of a difference between Si→SiGe and SiGe→Si interface widths indicates that Ge segregation is suppressed in these reduced-pressure chemical vapor deposition superlattices, so interface broadening is set by kinetically limited surface atomic exchange rather than by segregation.
  • Optical modeling of Si/SiGe heterostructures needs an interface layer with graded composition; the graded sigmoid model gives the lowest layer-thickness errors, while effective-medium approximations of the interface are inadequate.
  • Because the method works for layers as thin as 1.3 nm of Si and 2.2 nm of SiGe, it can be applied to gate-all-around nanosheet superlattices and other sub-10 nm multilayer devices where interface scattering limits performance.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the reported roughness and correlation lengths are physical, interface-roughness scattering models for Si/SiGe quantum cascade and nanosheet devices can be evaluated with measured inputs rather than fitted ones; a direct test would be to compute intersubband scattering rates from these values and compare them with measured linewidths.
  • The same iso-concentration-surface methodology should transfer to other epitaxial pairs, such as Si/SiC, GeSn/SiGeSn, or III–V superlattices, but the reconstruction calibration would need to be revalidated because evaporation-field contrasts differ between materials.
  • The apparent link between smaller roughness, larger correlation length, and lower growth temperature suggests a growth-front smoothing mechanism: at lower temperature the surface exchange length increases, so the roughness undulates more gently; this could be tested by growing a temperature staircase within a single superlattice and mapping the interfaces one by one.
  • Because spectroscopic ellipsometry consistently overestimates interface width relative to atom probe and electron energy loss data by less than 1 nm, a combined workflow could serve as a calibration transfer for non-destructive inline metrology of buried-interface sharpness.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This manuscript presents a method to map the roughness and lateral correlation of buried Si/SiGe interfaces in sub-10 nm superlattices using atom probe tomography (APT). The authors extract iso-concentration surfaces at 50% of the mean Ge concentration from APT reconstructions, compute the height-height correlation function H(τ), and fit it to extract RMS roughness σ and horizontal correlation length ξ. They report σ ≈ 0.2–0.3 nm and ξ ≈ 8.1–10.1 nm, with the 500 °C sample showing roughly 30% smaller σ and 24% larger ξ than the 650 °C samples. They also fit 1D Ge concentration profiles with a sigmoid to obtain interface widths of ~0.7–1.0 nm, and they compare four spectroscopic ellipsometry (SE) models (no interface, two effective-medium approximations, and a graded interface model) to argue that interfacial broadening must be included to describe the optical response.

Significance. The work addresses a real gap: buried interface roughness and correlation lengths in technologically relevant Si/SiGe heterostructures are presently treated as fit parameters in device and optical models, and direct 3D measurements are scarce. The multi-technique approach (APT, EELS, XTEM, HRXRD, SE) and the inclusion of multiple APT datasets per sample are strengths. If the quantitative claims hold, the results would provide useful input for modeling intersubband devices and Si nanosheet transistors. However, the central quantitative claims about σ and ξ rest on assumptions about APT reconstruction fidelity that are not fully validated, and the optical-modeling conclusion is logically narrower than the abstract states. The paper also includes a genuinely useful comparison of SE models with interface layers.

major comments (4)
  1. [Fig. 2(a), Eq. (1), and Section 'Extraction of RMS roughness and horizontal correlation length'] The reported RMS roughness values (σ ≈ 0.2–0.3 nm) are comparable to the height uncertainty expected from voxel counting statistics: with 1 nm³ voxels, ~25 at.% Ge, and typical APT detection efficiency, each voxel contains only tens of Ge atoms, so the iso-concentration surface height carries an uncertainty of order 0.1–0.2 nm. The manuscript does not quantify this noise floor or separate it from the physical roughness; the quoted ±11–13% uncertainties are standard deviations across interfaces and datasets, not measurement errors. Without a noise-floor estimate (e.g., from a Monte Carlo simulation of voxelated data or a reconstructed known-flat interface), the claim that σ varies from 0.2 to 0.3 nm cannot be distinguished from a constant value near the method's resolution limit.
  2. [Experimental section, 'Atom probe tomography', and SI Section 1] The APT reconstruction is calibrated only by iteratively matching layer thicknesses to XTEM within 5%, which constrains average vertical layer spacings but does not validate lateral geometry, local magnification, or sub-voxel height fluctuations. The SI reports that z-density correction did not change interface widths, but no test is presented for σ or ξ. The manuscript should either provide an independent validation of the roughness measurement (e.g., comparison with an established technique on a test structure) or explicitly discuss the systematic uncertainty this calibration leaves in the reported σ and ξ.
  3. [Fig. 2(d) and Table 1] The temperature dependence of σ and ξ is inferred from comparing S-16 and S-12 (650 °C), S-6 (600 °C), and S-3 (500 °C), but these samples differ not only in growth temperature but also in period count and layer thickness (e.g., S-3 has ~6–7.3 nm layers while S-16 has ~1.3–2.2 nm layers). The observed increase in ξ and decrease in σ at lower temperature could be influenced by thickness-dependent reconstruction artifacts or by the different number of interfaces averaged. The manuscript should acknowledge this confounding and, if possible, include samples grown at the same temperature with different periods to isolate the temperature effect.
  4. [SE studies, Fig. 5 and Table 2] The conclusion that 'observed atomic-level roughening at the interface must be accounted for' is not directly supported by the optical modeling, because the Mint_σ model parameterizes interfacial width (a graded composition profile) rather than the roughness parameters σ and ξ measured by APT. No optical model in the paper actually incorporates the APT-derived roughness or correlation length. The conclusion should be limited to interfacial broadening (width) until a model with explicit roughness parameters is tested.
minor comments (6)
  1. [Fig. 2(a) caption and text] The text refers to 'interface number 24' while the inset in the caption mentions 'interface 9'; this inconsistency should be corrected.
  2. [Abstract and Fig. 2(d)] The uncertainty intervals such as ±13% and ±11.5% should be explicitly defined as relative standard deviations of the mean across interfaces/datasets, not confidence intervals for the measured value.
  3. [Eq. (2) and SI Section 3.2] The relationship between the sigmoid parameter ℒ and the reported interface width is only stated in the SI as dint^(i) = 4ℒ; this definition should appear in the main text where Eq. (2) is introduced.
  4. [Table 1] Table 1 is difficult to parse: the columns for XTEM, HRXRD, and SE thicknesses are not clearly aligned, and the 'surface rms. roughness' values are interleaved with the thickness data; reformatting is needed.
  5. [Section 'Interfacial width' (discussion of Ge segregation)] The statement that 'there is no Ge segregation' is stronger than the data support, since the rising and falling interface widths are compared at the resolution limit of the measurement; consider softening this claim.
  6. [Fig. 2(a) and Eq. (1)] The fitted Hurst parameter α is never reported, even though it is a free parameter in H_fit(τ) and influences the relationship between σ and ξ; reporting α for at least the representative interfaces would aid reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: roughness and correlation values are direct APT data analyses, and the SE claim is cross-validated against independent XTEM/APT inputs rather than derived from them.

full rationale

The central quantitative claims are not predictions derived from fitted inputs. The RMS roughness σ and horizontal correlation length ξ are extracted by applying Eq. (1) and the standard fit Hfit(τ)=2σ²[1−exp(−(τ/ξ)^(2α))] to APT-generated iso-concentration surfaces, so the reported values are direct analyses of the measured atom maps rather than outputs of a model whose parameters were fit to those same values. The APT reconstruction is calibrated to XTEM only through average layer thickness: the paper states that reconstruction was done 'iteratively until the layer thickness of the 3-D APT reconstruction matches (with 5.0 % tolerance) that obtained from the corresponding XTEM image.' This constrains average vertical spacings and does not inject the target σ or ξ values into the analysis. The ellipsometry portion is also non-circular: the SI explicitly says, 'To avoid observer-bias, input related to interface width, layer thickness, and composition from XTEM and APT was not been used to develop the interface-related optical model,' with HRXRD supplying the initial thicknesses and XTEM/APT used afterward only to verify the model. The conclusion that the graded Mintσ model is needed is based on external figures of merit (MSE and ΔErr relative to XTEM), not on a parameter that was defined as the conclusion. The fact that Mintσ has more floating parameters and therefore may be favored is a model-selection risk, not a circular reduction. The self-reference [31] is a generic methodological citation for APT and carries no uniqueness theorem or fitted value. The acknowledged APT reconstruction artifacts and the possible voxel-statistics floor are measurement-fidelity concerns, not circularity, because no equation in the paper defines an output quantity as an input quantity. No step meets the evidentiary standard for a circularity finding.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper's quantitative outputs are obtained by fitting standard statistical models to APT and SE data. The main assumptions concern APT reconstruction fidelity and the phenomenological functions used to define interfaces and extract roughness; no new physical entities are introduced.

free parameters (6)
  • RMS roughness sigma per interface (Hfit parameter) = 0.2-0.3 nm; per-interface values in Fig. 2(b), Fig. S4, Fig. S5
    Fitted to height-height correlation Hfit(tau) = 2 sigma^2 [1 - exp(-(tau/xi)^(2 alpha))]; reported as mean +/- standard deviation.
  • Horizontal correlation length xi per interface (Hfit parameter) = 8.1 nm (S-16), 8.4 nm (S-6), 10.1 nm (S-3)
    Second fitting parameter of Hfit; the crossover distance where H(tau) saturates.
  • Hurst parameter alpha per interface = not reported
    Third parameter in Hfit; fit quality R^2 > 0.99 is reported but alpha values are not listed.
  • Sigmoid interface-width parameter L (or derived width d_int) = 1.0 nm (S-16), 0.9 nm (S-6), 0.7 nm (S-3)
    Fitted to 1D Ge concentration profiles with c(x) = c0 + d0 / [1 + exp(-(x0 +/- x)/L)]; interface width is derived from L.
  • SE interface width d_int and Ge content d0 in Mint-sigma model = d_int about 1.3-1.8 nm; d0 about 12-17 at.% Ge
    Floating parameters in the graded alloy optical model, iteratively fitted to ellipsometry data.
  • EMA fraction in EMA-based interface models = <10 at.% (Mint_1) or >50 at.% (Mint_2)
    Fitted to ellipsometry data; the large discrepancy between models is used to argue against EMA interfaces.
assumptions (4)
  • domain assumption APT reconstruction with hemispherical evaporation and parameters calibrated to XTEM thickness within 5% gives correct 3D geometry.
    Invoked around Figure 1(c) and the experimental section; if the reconstruction geometry is distorted, the height statistics and iso-surfaces would be artifacts.
  • domain assumption The Ge iso-concentration surface at 50% of the mean layer Ge content represents the physical interface.
    All roughness and correlation analyses are performed on these surfaces (Figure 1(d)); the authors note that intermixing increases polygon-placement uncertainty.
  • domain assumption The height-height correlation function Hfit = 2 sigma^2 [1 - exp(-(tau/xi)^(2 alpha))] adequately describes buried interface morphology.
    Used to extract sigma and xi; R^2 > 0.99 is stated, but no residual analysis or alternative correlation models are shown.
  • domain assumption Bulk diffusion is negligible during growth; interfacial width is set by a kinetically limited surface atomic exchange process.
    This is the mechanistic explanation for lower-temperature sharpening, presented near Figure 3(c) and supported by refs [21-23].

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3-D Atomic Mapping of Interfacial Roughness and its Spatial Correlation Length in sub-10 nm Superlattices." pith.science (2026). https://pith.science/paper/F45UTPBN

@misc{pith2026190800874,
  author       = {Pith},
  title        = {Pith review of: 3-D Atomic Mapping of Interfacial Roughness and its Spatial Correlation Length in sub-10 nm Superlattices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F45UTPBN}},
  note         = {Machine review of arXiv:1908.00874}
}
read the original abstract

The interfacial abruptness and uniformity in heterostructures are critical to control their electronic and optical properties. With this perspective, this work demonstrates the 3-D atomistic-level mapping of the roughness and uniformity of buried epitaxial interfaces in Si/SiGe superlattices with a layer thickness in the 1.5-7.5 nm range. Herein, 3-D atom-by-atom maps were acquired and processed to generate iso-concentration surfaces highlighting local fluctuations in content at each interface. These generated surfaces were subsequently utilized to map the interfacial roughness and its spatial correlation length. The analysis revealed that the root mean squared roughness of the buried interfaces in the investigated superlattices is sensitive to the growth temperature with a value varying from about 0.2 nm (+/- 13%) to about 0.3 nm (+/- 11.5%) in the temperature range of 500-650 Celsius. The estimated horizontal correlation lengths were found to be 8.1 nm (+/- 5.8%) at 650 Celsius and 10.1 nm (+/- 6.2%) at 500 Celsius. Additionally, reducing the growth temperature was found to improve the interfacial abruptness, with 30 % smaller interfacial width is obtained at 500 Celsius. This behavior is attributed to the thermally activated atomic exchange at the surface during the heteroepitaxy. Finally, by testing different optical models with increasing levels of interfacial complexity, it is demonstrated that the observed atomic-level roughening at the interface must be accounted for to accurately describe the optical response of Si/SiGe heterostructures.

Figures

Figures reproduced from arXiv: 1908.00874 by the authors.

Figure 1
Figure 1. XTEM and APT 3-D reconstruction of sample S-16. (a) XTEM image of sample S￾16. A zoom into the selected regions of the image is shown alongside using the numbered boxes. All scale bars for the images within these boxes correspond to 5 nm. (b) 3-D atom-by￾atom APT reconstruction of sample S-16. (c) Top: 3-D reconstruction of the atoms lying within the black cylinder (diameter 30 nm) at the center of the 3D reconstruc… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

  1. [1]

    Vurpillot, D

    F. Vurpillot, D. Larson, A. Cerezo, Surf. Interface Anal. 2004, 36, 552

    work page 2004

  2. [2]

    O. Dyck, D. N. Leonard, L. F. Edge, C. A. Jackson, E. J. Pritchett, P. W. Deelman, J. D. Poplawsky, Adv. Mater. Interfaces 2017, 4, 1700622

    work page 2017

  3. [3]

    J. G. Brons, A. A. Herzing, K. T. Henry, I. M. Anderson, G. B. Thompson, Thin Solid Films 2014, 551, 61

    work page 2014

  4. [4]

    H. G. Tompkins, E. A. Irene, I. An, H. Arwin, C. Chen, R. W. Collins, A. S. Ferlauto, J. N. Hilfiker, J. Humlícek, J. Gerald E. Jellison, J. Lee, F. A. Modine, A. Röseler, M. Schubert, J. A. Zapien, Handbooks of Ellipsometry, William Andrew Publishing And Springer-Verlag GmbH & Co. KG, 2005

    work page 2005

  5. [5]

    Fujiwara, Spectroscopic Ellipsometry Principles and Applications, John Wiley & Sons, Ltd, 2003

    H. Fujiwara, Spectroscopic Ellipsometry Principles and Applications, John Wiley & Sons, Ltd, 2003

    work page 2003

  6. [6]

    T. C. Asmara, I. Santoso, A. Rusydi, Rev. Sci. Instrum. 2014, 85, 123116

    work page 2014

  7. [7]

    D. E. Aspnes, Surf. Sci. 1980, 101, 84

    work page 1980

  8. [8]

    D. E. Aspnes, Thin Solid Films 1982, 89, 249

    work page 1982

Show all 15 references
  1. [9]

    J. L. Freeouf, Appl. Phys. Lett. 1988, 53, 2426. 2 µm 46

    1988

  2. [10]

    V Nguyen, J

    N. V Nguyen, J. G. Pellegrino, P. M. Amirtharaj, D. G. Seiler, S. B. Qadri, J. Appl. Phys. 1993, 73, 7739

    1993

  3. [11]

    T. H. Ghong, Y. D. Kim, D. E. Aspnes, M. V. Klein, D. S. Ko, Y. W. Kim, V. Elarde, J. Coleman, J. Korean Phys. Soc. 2006, 48, 1601

    2006

  4. [12]

    M. Born, E. Wolf, Principles of Optics, Cambridge University Press, Cambridge MA, 1999

    1999

  5. [13]

    Aspnes, A

    D. Aspnes, A. Studna, Phys. Rev. B 1983, 27, 985

    1983

  6. [14]

    G. E. J. Jr., Opt. Mater. (Amst). 1992, 1, 41

    1992

  7. [15]

    C. M. Herzinger, B. Johs, W. A. McGahan, J. A. Woollam, W. Paulson, J. Appl. Phys. 1998, 83, 3323

    1998

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.