Tilt-Corrected Bright Field (tcBF) Imaging

• tcBF is a bright-field imaging strategy in 4D-STEM that compensates for tilt-dependent misregistration to produce direct phase-contrast images.
• It employs virtual off-axis detectors by shifting and summing defocus-induced tilt-corrected images, thereby reducing blurring and geometric distortion.
• By integrating the full bright-field disk, tcBF achieves an information limit up to 2α and improves dose efficiency, offering enhanced imaging of fine structural details.

Tilt-corrected bright field (tcBF) is a bright-field imaging strategy in which tilt-dependent misregistration is explicitly compensated before signals are combined, so that coherent phase contrast adds constructively rather than being blurred or geometrically distorted. In its most precise current usage, tcBF is a 4D-STEM imaging mode where each bright-field-disk pixel is treated as a virtual off-axis bright-field detector, its image is shifted back by the defocus-induced tilt shift $\Delta \mathbf{p} = \Delta f\,\boldsymbol{\theta}$, and all such shift-corrected images are summed (Ma et al., 28 Jul 2025). Under the weak phase object approximation, this produces a direct phase-contrast bright-field image with an information limit up to $2\alpha$, and it forms one branch of a broader tilt-corrected 4D-STEM framework that also includes tcDPC and acBF (Ma et al., 1 Oct 2025).

1. Definition, reciprocity, and scope

In 4D-STEM, a convergent probe is scanned across the specimen and a 2D convergent-beam electron diffraction pattern is recorded at each probe position. The central disc of this pattern—the region within the probe semi-convergence angle $\alpha$—is the bright-field disk. Each detector pixel inside that disk can be treated as a tiny off-axis bright-field detector. By reciprocity, a STEM image formed from one pixel at scattering angle $\boldsymbol{\theta}$ corresponds to a CTEM image formed with a plane-wave incident at tilt $\boldsymbol{\theta}$ (Ma et al., 28 Jul 2025).

Within Rose’s generalized contrast formalism, the elastic scattering amplitude is decomposed as

$F = F_s + iF_a,$

with

$$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$

Under the weak phase object approximation, tcBF is obtained by summing the symmetric components of the complex phase contrast transfer function across the bright-field disk after tilt correction, whereas tcDPC is obtained by summing the antisymmetric components after tilt correction (Ma et al., 28 Jul 2025).

The term is used most strictly in this 4D-STEM sense. A broader interpretation appears in adjacent literatures on bright-field and annular bright-field STEM, full-field TEM with tilt illuminations, and scanned optical systems, where “tilt correction” addresses specimen mistilt, defocus-induced parallax, illumination-plane tilt, or background nonuniformity. This suggests a family of methods organized around controlling or compensating tilt before image integration or interpretation.

2. 4D-STEM image formation and the tcBF construction

For pure defocus, the aberration phase is

$$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$

where $k_0 = 2\pi/\lambda$ is the electron wave number and $\Delta f$ is defocus. The phase-contrast analysis shows that the image formed from detector angle $2\alpha$0 is shifted in real space by

$2\alpha$1

This shift blurs any bright-field image that simply sums all detector pixels, because each pixel images the same object but slightly shifted (Ma et al., 28 Jul 2025).

Under the weak phase object approximation, the phase-contrast contribution for a point-like detector at $2\alpha$2 can be written in Fourier space as

$2\alpha$3

with

$2\alpha$4

For defocus only, this becomes

$2\alpha$5

The final exponential is the Fourier-shift factor corresponding to the real-space displacement $2\alpha$6 (Ma et al., 28 Jul 2025).

The operational tcBF algorithm is correspondingly direct. First, define the bright-field disk by choosing detector pixels with $2\alpha$7. Second, form one virtual STEM image per detector pixel: $2\alpha$8 Third, determine defocus $2\alpha$9, either from microscope settings or from cross-correlation of the off-axis images. Fourth, shift each image by

$\alpha$0

implemented through a Fourier shift,

$\alpha$1

Finally, sum the shift-corrected images over all pixels within the bright-field disk: $\alpha$2 In practice, upsampling the scan grid before shifting allows sub-pixel shifts and finer real-space sampling than the native scan pitch (Ma et al., 28 Jul 2025).

3. Transfer functions, information limit, and neighboring modalities

The defining transfer property of tcBF is obtained by summing conjugate detector pairs after tilt correction. Under pure defocus, the tcBF phase-contrast transfer function is

$\alpha$3

where $\alpha$4, and $\alpha$5 is the normalized overlap area between two disks of radius $\alpha$6 separated by distance $\alpha$7: $\alpha$8 The information limit is therefore $\alpha$9, twice the axial bright-field STEM limit $\boldsymbol{\theta}$0, while the sinusoidal modulation remains CTEM-like, with oscillatory contrast transfer and zero crossings (Ma et al., 28 Jul 2025).

The associated tilt-corrected modes are complementary rather than redundant.

Mode	Combination rule	Transfer characteristic
tcBF	Sum of shift-corrected bright-field-disk images; symmetric combination of $\boldsymbol{\theta}$1	$\boldsymbol{\theta}$2
tcDPC	Antisymmetric difference of conjugate detector pairs after tilt correction	$\boldsymbol{\theta}$3
acBF	Fourier-space combination of symmetric and antisymmetric tilt-corrected channels	$\boldsymbol{\theta}$4

The tcDPC transfer function is

$\boldsymbol{\theta}$5

and is purely imaginary, yielding a differential or gradient-phase signal in real space. It has no sensitivity to uniform phase at $\boldsymbol{\theta}$6 and is best at in-focus conditions. The integrated form i-tcDPC is obtained by division by $\boldsymbol{\theta}$7, but does not increase information content (Ma et al., 28 Jul 2025).

acBF combines the tcBF and tcDPC channels so that the total contrast transfer is continuously nonzero from $\boldsymbol{\theta}$8 to $\boldsymbol{\theta}$9. In the terminology of the 4D-STEM literature, tcBF mainly uses symmetric triple-overlap information, tcDPC uses antisymmetric double-overlap information, and acBF combines both coherently. This is why acBF is described as making maximal use of bright-field-disk information under the weak phase object approximation, whereas tcBF alone remains a direct phase-contrast mode with CTEM-like oscillations (Ma et al., 28 Jul 2025).

These relationships also locate tcBF within direct ptychography. Direct ptychography methods such as single-sideband and iDPC mainly use double-overlap sidebands. tcBF uses the symmetric triple-overlap channel, tcDPC uses the antisymmetric double-overlap channel, and acBF combines both. Iterative multislice ptychography can additionally exploit coherent amplitude terms and dark-field information. tcBF is therefore a fast, non-iterative subset of the information exploited by full ptychography.

4. Higher-order aberrations, Scherzer operation, and dose efficiency

The clean defocus-only picture does not extend unchanged to higher-order aberrations. In the aberration-aware 4D-STEM treatment, the probe-forming lens phase is written as

$\boldsymbol{\theta}$0

and the real-space shift of a virtual bright-field image is, to first order,

$\boldsymbol{\theta}$1

For defocus alone, this reduces to $\boldsymbol{\theta}$2. For higher-order aberrations, however, there are additional tilt-dependent changes in the effective aberration function of each off-axis virtual bright-field image, which tcBF does not correct (Ma et al., 1 Oct 2025).

Spherical aberration is the canonical example. With defocus $\boldsymbol{\theta}$3 and spherical aberration $\boldsymbol{\theta}$4, each detector pixel acquires a different effective defocus and a tilt-dependent effective astigmatism. If the images are simply shift-corrected and summed, the oscillatory CTFs from different $\boldsymbol{\theta}$5 add out of phase, strongly damping the overall tcBF transfer, particularly at low frequencies. This is why tcBF is described as fragile to higher-order aberrations, whereas acBF computes and corrects the complete complex PCTF per virtual image before summation (Ma et al., 1 Oct 2025).

The same work emphasizes that aberrations need not only be suppressed; they can be used. At Scherzer defocus in a spherically-aberration-limited system, the resultant phase shift from the probe-forming lens acts as a phase plate, removing oscillations from the acBF CTF. The cited Scherzer condition is

$\boldsymbol{\theta}$6

In this regime, acBF becomes a non-iterative bright-field phase-contrast method with continuously nonzero transfer up to $\boldsymbol{\theta}$7, even in the presence of higher-order aberrations (Ma et al., 1 Oct 2025).

The same aberration-aware framework is tied directly to dose efficiency. The detective quantum efficiency is written as

$\boldsymbol{\theta}$8

For an ideal pixel detector with flat Poisson-limited noise, DQE follows the squared transfer magnitude. Because axial bright-field collects only a small fraction of the bright-field disk, it discards most forward-scattered electrons. tcBF and acBF integrate over the whole bright-field disk, but tcBF loses performance when aberrations damp the transfer. In the reported simulations, a NU-1000 MOF dataset at $\boldsymbol{\theta}$9, $F = F_s + iF_a,$0, $F = F_s + iF_a,$1, $F = F_s + iF_a,$2, and Scherzer defocus $F = F_s + iF_a,$3 shows axial bright field dominated by noise, tcBF improving large features but failing to resolve atomic columns, and acBF resolving individual Zr atoms and parts of the organic linker at low dose (Ma et al., 1 Oct 2025).

5. Specimen mistilt and quantitative bright-field/ABF metrology

A distinct but closely related use of tilt correction appears in annular bright-field and bright-field STEM metrology, where the issue is not detector-pixel parallax but specimen mistilt. In SrTiO$F = F_s + iF_a,$4, a small specimen tilt of $F = F_s + iF_a,$5 in a $F = F_s + iF_a,$6 thick sample along $F = F_s + iF_a,$7 causes an artificial displacement between O and Sr/TiO columns of $F = F_s + iF_a,$8, more than 3 times the scan-noise and sample-drift induced image distortion of $F = F_s + iF_a,$9. The artifact depends on crystal mistilt angle, specimen thickness, defocus, convergence angle, and uncorrected aberration (Gao et al., 2017).

The physical origin is asymmetric channeling and dechanneling. Heavy cation columns have strong channeling and comparatively robust measured positions, whereas light anion columns have weak channeling and their contrast is strongly influenced by dechanneling electrons from nearby heavy columns. In ABF and, by extension, central bright-field STEM, this makes measured O positions far more tilt-sensitive than heavy-column positions. The cited simulations show that at $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$0 the artificial displacement can reach $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$1, while cation-cation sublattice measurements remain comparatively stable (Gao et al., 2017).

To separate tilt from drift and scan distortion, the paper introduces a relative distance measurement. For a column of type $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$2 and its four nearest neighbors of type $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$3,

$$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$4

Under slowly varying drift, the drift contribution approximately cancels in $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$5. This provides a local mixed-sublattice displacement vector that can reveal systematic tilt fields even when cation-cation metrics look regular. The reported practical threshold is that, for a $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$6 SrTiO$$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$7 specimen at $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$8 with $$F_s(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) + F^*(\mathbf{k},\mathbf{k}')}{2},\quad F_a(\mathbf{k}',\mathbf{k}) = \frac{F(\mathbf{k}',\mathbf{k}) - F^*(\mathbf{k},\mathbf{k}')}{2i}.$$9 convergence, keeping residual tilt below $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$0 keeps O-versus-cation artifacts $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$1, whereas above $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$2–$$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$3 cation-anion displacements become $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$4 and dominate over noise (Gao et al., 2017).

This is not the same definition of tcBF as in 4D-STEM, but it addresses the same general problem: image formation is altered by tilt-dependent asymmetry, and quantitative bright-field interpretation requires either hardware alignment, simulation-based calibration, or explicit post hoc correction.

6. Optical and full-field analogues

A broader geometric interpretation of tcBF appears in scanned optical systems. In the oblique-plane microscopy literature, tilt-corrected bright-field imaging is described as, at its core, a problem of controlling and/or compensating the tilt of an illumination/detection plane as it is scanned. For an ideal thin lens of focal length $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$5, a point at lateral offset $$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$6 in the back focal plane corresponds to a collimated beam with tilt

$$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$7

Strict tilt invariance requires that the scanner rotation axis lie in the back focal plane and that the incident beam pass through the rotation axis, i.e.

$$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$8

In the optimized geometry, the planar mirror scan adds no additional aberrations to an otherwise aberration-free lens, and the central design rule is that the pivot point of the beam in the back focal plane remain fixed as the mirror scans (Kumar et al., 2020).

A full-field TEM analogue appears in high-resolution imaging of 3D nanocrystals with tilt illuminations. There, a strongly tilted illumination is combined with a ring slit placed immediately behind the sample inside the objective pole-piece. Only Bragg-diffracted beams on the selected ring are transmitted, while the direct beam also passes the ring slit and acts as a reference wave. The proposed geometry is explicitly contrasted with conventional TEM: the author argues that conventional CTF mixing in 3D crystals arises from spherical-wave phase contrast, whereas in the new microscopy there are no spherical waves, so the CTF is not observed in principle. The resulting images are described as similar to those obtained using STEM and as projection-like “volume holograms” of the crystal. This can be interpreted as a tilt-engineered bright-field TEM mode in a broad tcBF sense, although the paper does not itself use the term tcBF (Shintake, 6 Jun 2026).

A third analogue appears in quantitative transmission bright-field light microscopy. A spectroscopic correction pipeline maps raw counts to radiant flux on a pixel-by-pixel basis,

$$\chi(\boldsymbol{\theta}) = \frac{1}{2} k_0 \Delta f |\boldsymbol{\theta}|^2,$$9

so that corrected intensities are proportional to photon flux. The corrected blank field becomes nearly uniform, and dust, vignetting, tap imbalance, and channel-dependent response are normalized out. This does not perform angular or phase-space tilt correction, but it provides a tcBF-like flat-field foundation in which residual spatial variation is more directly attributable to the sample (Platonova et al., 2019).

7. Assumptions, limits, and interpretive boundaries

The cleanest tcBF theory is explicitly tied to the weak phase object approximation. Under WPOA, the sample transmission is written as

$k_0 = 2\pi/\lambda$0

and the tcBF image obeys the linear model

$k_0 = 2\pi/\lambda$1

Beyond WPOA, coherent amplitude terms and quadratic terms from inelastic or plural scattering become important. The 4D-STEM literature therefore treats tcBF as a direct phase-contrast mode rather than a complete scattering reconstruction, and identifies additional usable information in tcDF and iterative ptychography that tcBF does not use (Ma et al., 28 Jul 2025).

Detector and sampling assumptions also matter. The 4D-STEM formulation assumes sufficiently fine detector sampling within the bright-field disk so that each pixel samples a small angular region and the plane-wave approximation per pixel remains valid. The aberration-aware extension assumes that the shift field and the aberration function can be estimated accurately enough to correct each virtual image. Non-round aberrations such as two-fold astigmatism can create true anisotropic information loss, so acBF cannot recover information that is absent from the dataset (Ma et al., 1 Oct 2025).

In the optical analogues, the governing approximations differ but are equally restrictive. The scanned-oblique-plane treatment is derived in geometrical optics with ideal thin-lens behavior, small-angle approximations, and a perfectly flat mirror. Real objectives, wave effects, and mechanical tolerances can reintroduce tilt variance. In the ABF-STEM metrology setting, correction depends on thickness, defocus, convergence angle, and material-specific multislice behavior rather than on a universal linear transfer law. These distinctions matter because “tilt-corrected bright field” names a common operational logic—realigning or compensating tilt before interpretation—but not a single universal forward model (Kumar et al., 2020).

Taken narrowly, tcBF denotes the symmetric, defocus-corrected bright-field reconstruction from 4D-STEM. Taken more broadly, it identifies a class of bright-field methods in which tilt is treated as a controllable or calibratable degree of freedom rather than as an unmodeled nuisance. The modern literature places its most rigorous formulation in 4D-STEM, where tcBF, tcDPC, acBF, and related modes supply a continuum between conventional bright-field imaging and direct ptychography.