Virtual Cornea Addition Cannot Substitute for a Physical Cornea in IOL Wavefront Measurement: Evidence from Optical Simulation

Virtual Cornea Addition Cannot Substitute for a Physical Cornea in IOL Wavefront Measurement: Evidence from Optical Simulation

Yiska Fattal R&D Physicist, Rotlex

June 2026     |     Software: Zemax OpticStudio     |     Standard reference: ISO 11979-2 

ABSTRACT

ISO 11979-2 specifies that intraocular lens (IOL) optical performance shall be evaluated in a model eye that includes a model cornea. A suggested simplification is to measure the IOL wavefront in isolation and then add the wavefront contribution of an idealized cornea. Using ray-tracing simulations in Zemax OpticStudio, we demonstrate that two IOL designs that are optically identical when measured alone – producing the same wavefront error and power profile under collimated illumination – produce substantially different system-level wavefronts once a model cornea is placed anterior to them. The divergence arises because each IOL surface refracts rays according to the local angle of incidence, which depends on the curvature and aberration content of the incoming wavefront. A converging beam produced by the cornea impinges each IOL surface with a different angle-of-incidence distribution than a collimated beam does; two lens geometries that happen to produce the same output under collimated illumination will generally not produce the same output under converging illumination. Virtual post-processing addition of a corneal wavefront cannot reproduce this input-dependent refraction and therefore cannot satisfy the conditions for ISO-compliant MTF measurement.

1. INTRODUCTION

The ISO 11979-2 standard for the optical characterization of intraocular lenses requires that modulation transfer function (MTF) measurements be performed in a model eye incorporating a model cornea with defined paraxial power and spherical aberration (optional). The intent is to replicate, at least in a simplified form, the aberration environment the implanted IOL will encounter within the human eye.

In practice, some measurement protocols propose to characterize the IOL wavefront in isolation – illuminated by a collimated beam without a cornea present – and subsequently to add the known aberration of a wanted model cornea to the isolated lens measured wavefront in software. The appeal of this approach is practical: a single IOL measurement suffices, and the corneal contribution can be changed in post-processing without remeasuring the lens. The assumption it requires, however, is that wavefront aberrations are additive: if the IOL alone introduces wavefront W₂ and the cornea alone introduces W₁, then the combination produces Wsystem= W₁ + W₂.

This note challenges that assumption by demonstrating through ray-trace simulation that two IOL designs with nominally identical isolated wavefronts produce markedly different combined wavefronts once a physically present model cornea is placed in the system.

1.1 The physical reason: angle-of-incidence dependence on input wavefront curvature

A critical geometric fact shapes this problem: in a well-designed measurement system, the aperture stop of the optical train is placed at or near the first (anterior) surface of the IOL. This means that the marginal ray height at the IOL is determined entirely by the stop size and is, by definition, identical whether or not a cornea is present upstream. The cornea does not change where rays land on the IOL; it changes the direction in which they arrive.

Core principle: With the aperture stop at the IOL, every ray that enters the system reaches the IOL at the same height regardless of what optics precede the stop. What the cornea changes is the angle of incidence of each ray at every IOL surface – and it is these angles, not the ray heights, that determine how much aberration each surface introduces.

Snell’s law at a refracting surface produces an output ray direction that is a nonlinear function of the input ray direction and the local surface normal. For a collimated input beam, all rays arrive at the IOL’s first surface travelling parallel to the optical axis: the angle of incidence at each radial zone is determined solely by that surface’s local slope at that height. For a converging beam produced by the cornea, each ray arrives already tilted toward the focal point of the cornea; the angle of incidence at the same radial zone is therefore the sum of the surface slope and the ray convergence angle. The nonlinearity of Snell’s law means that the aberration introduced by the surface under these two illumination conditions is not simply offset by a constant.

Furthermore, the cornea can introduce its own wavefront aberration, according to the options in ISO 11979-2 annex B. The beam arriving at the IOL is therefore not a clean converging cone but a converging beam with a non-flat wavefront. The IOL surfaces interact with this structured input wavefront in a way that depends on the detailed surface geometry: two IOL shapes that produce the same output from a flat (collimated) input wavefront will, in general, produce different outputs from a curved, aberrated, input wavefront. The mapping from input pupil angle to output wavefront error is geometry-specific and cannot be characterized by a single scalar measurement under one illumination condition.

This is not an approximation error or a second-order effect. It is a fundamental consequence of Snell’s law being nonlinear in the ray angles. Wavefront addition – W₁ + W₂,collimated – would be exactly correct only if both elements were thin phase elements at the same plane (a paraxial system). The simulation below makes this argument concrete by constructing a pair of IOLs that are identical under collimated illumination but diverge substantially once the cornea is present.

2. LENS DESIGNS AND SYSTEM CONFIGURATION

Two IOL designs were constructed in Zemax OpticStudio, each simulated in two conditions: the IOL illuminated by a collimated beam alone (lens-only), and the same IOL preceded by an aberration-free model cornea (lens-with-cornea).

2.1 Configuration 2: Plano-convex lens

The first design is a plano-convex element with a flat posterior surface and all refractive power concentrated on the anterior convex surface (radius 6.5 mm). No conic constant is used. 

2.2 Configuration 1: Biconvex lens with conic constant

The second design is a Bi-convecx IOL with optical power distributed across two curved interfaces. A conic constant on the anterior surface introduces controlled asphericity. Both anterior and posterior radius are 12.97 mm.

Both designs are at typical IOL conditions – 0.2mm ET and positioned in a wet cell – and were tuned so that their isolated wavefront errors under collimated illumination are as close as possible. The Aperture stop size at both designs is 5.15mm and located on the anterior surface.

3. OPTICAL LAYOUT

Figure 2 shows the 2-D ray-trace layouts for each design in isolation under collimated illumination. Both lenses bring the beam to a paraxial focus at comparable distances, and the layouts appear qualitatively similar. The stop at the IOL ensures the same beam footprint at that plane in all configurations.

Figure 2. Zemax layout diagrams of the two IOL configurations in a wet cell and under collimated illumination (no cornea). Left: Configuration 1 (plano-convex); right: Configuration 2 (biconvex, conic). Total axial lengths: 59.38 mm and 59.74 mm. The input beam is collimated in both cases, meaning all rays arrive at the IOL parallel to the optical axis.

Figure 3 shows the same two designs with the model cornea inserted. The cornea converts the collimated input into a converging beam. Because the stop remains at the IOL, ray heights there are unchanged – but every ray now arrives at the IOL with a non-zero convergence angle whose magnitude depends on its radial position. This altered angle-of-incidence distribution is what each IOL geometry “sees” differently.

Figure 3. Zemax layout diagrams with the model cornea present (left: Config. 1, axial length 29.62 mm; right: Config. 2, axial length 29.78 mm). The ray heights at the IOL are fixed by the stop and are identical to Figure 2, but every ray now arrives at the IOL with a non-zero convergence angle whose magnitude depends on its radial position. The two lens geometries respond to this structured illumination differently.

4. POWER PROFILES

The Power Pupil Map (Y-pupil scan, spherical power in diopters at λ = 0.546 μm, on-axis field) shows how effective power varies with normalized pupil height – a direct signature of the spherical aberration structure of each system.

4.1 Lens-only power profiles (collimated illumination)

Figure 4. Power Pupil Map for Config. 1 (Pcvx) and Config. 2 (Bicvx) under collimated illumination, without cornea. Both profiles span approximately 19–22.1 D and trace a nearly identical U-shaped curve. The two designs are optically equivalent under this illumination condition.

4.2 Power profiles with cornea present (converging illumination)

Figure 5. Power Pupil Map for Config. 1 (Pcvx) and Config. 2 (Bicvx) with the model cornea present. Config. 1 spans approximately 46.5–47.3 D; Config. 2 spans 46.4–48.5 D. The amplitude of power variation across the pupil is roughly 2.5× larger in Config. 2. Two designs that appeared identical now exhibit clearly different aberration structures.

Key observation: Under collimated illumination, the peak-to-valley power variation of both designs is less than 0.3 D and the profiles are virtually superimposable. Once the cornea is inserted, the peak-to-valley power variation becomes approximately 0.8 D for Config. 1 and approximately 2 D for Config. 2 – a factor of 2.5 difference between two lenses that appeared optically identical. The stop position was unchanged; only the wavefront curvature of the incident beam changed.

5. ZERNIKE WAVEFRONT ANALYSIS

Wavefront error was decomposed onto Zernike Standard polynomials at 5.15mm aperture diameter. The RMS wavefront error (to centroid), peak-to-valley error, and key Zernike coefficients are presented for both measurement conditions.

5.1 Lens-only Zernike analysis (collimated illumination)

Figure 6. Zernike Standard Coefficient reports for Config. 1 (upper) and Config. 2 (lower) without cornea. The dominant terms are piston (Z1) and primary spherical aberration (Z11 ≈ 0.327 waves in both). RMS to centroid: 1.246 waves in both cases. Under collimated illumination the two designs are indistinguishable.

5.2 Zernike analysis with cornea present (converging illumination)

Figure 7. Zernike Standard Coefficient reports for Config. 1 (upper) and Config. 2 (lower) with the model cornea present. Config. 1: RMS (centroid) = 0.710 waves, Z11 = 0.174 waves. Config. 2: RMS (centroid) = 1.903 waves, Z11 = 0.493 waves. Secondary spherical aberration (Z22) also differs substantially. These differences are entirely absent from the collimated-beam measurements.

6. THROUGH-FOCUS MTF RESPONSE

The resulting Through-Focus Response (TFR) represents the most clinically and regulatorily relevant metric for IOL quality evaluation: it shows the modulus of the optical transfer function (MTF) at a fixed spatial frequency as a function of axial focus shift. Crucially, ISO 11979-2 mandates that this measurement be performed in the cornea-plus-lens configuration, not on the IOL in isolation. The TFR therefore provides a direct, standard-compliant demonstration of the performance difference between the two lens designs.

Figure 8. Polychromatic Through-Focus MTF at 100 lpm, 3 mm aperture (λ = 0.546 μm), on-axis field, cornea-plus-lens configuration. Configuration 1 (Pcvx): peak MTF ≈ 0.64. Configuration 2 (bicvx): peak MTF ≈ 0.57. The peak MTF values differ by approximately 11% despite both lenses being optically identical when measured in isolation.

6.1 Interpretation of the TFR differences

The most striking feature of Figure 8 is the difference in peak MTF: Configuration 1 reaches approximately 0.64 while Configuration 2 reaches approximately 0.57 – an 11% reduction at 100 lp/mm. This difference is well above measurement noise and would be clearly distinguishable on a standard ISO 11979-2 bench. The degraded MTF in Configuration 2 is a direct consequence of the larger spherical aberration that its single high-curvature surface accumulates when illuminated by the converging corneal beam, as quantified in Sections 4 and 5.

This result would not be predicted by adding a corneal wavefront to the isolated IOL measurement. The virtual-addition approach would assign the same peak MTF to both configurations, because their isolated wavefronts are identical. Only a system-level measurement with the cornea physically present correctly predicts the MTF divergence shown in Figure 8.

ISO 11979-2 relevance: The standard specifies a minimum MTF value that must be met at defined spatial frequencies in the cornea-plus-lens configuration. Figure 8 shows that the two designs, indistinguishable in isolation, would yield different pass/fail outcomes if the MTF threshold lies between 0.57 and 0.64. A virtual-cornea approach, by contrast, would report identical MTF for both and could therefore incorrectly certify a non-compliant design.

Table 1. Summary of key optical metrics for both IOL configurations under both measurement conditions. Metrics marked – are not applicable in the lens-only configuration per ISO 11979-2.

7. DISCUSSION

7.1 Why wavefront addition fails: the angle-of-incidence argument

The wavefront of a system with two sequential optical elements is not, in general, the sum of the wavefronts those elements produce individually. Wavefront addition holds exactly only when each element can be treated as a thin phase screen at the same plane, so that the ray bundle is identical at each element whether or not the other is present. In all other configurations, the aberration contribution of the second element depends on the shape of the wavefront presented to it by the first.

The relevant quantity at each refracting surface is the angle of incidence of each ray. Snell’s law is nonlinear in angles: the aberration a surface introduces is not simply proportional to the ray height but depends on the angle at which each ray meets the local surface normal. When the IOL is illuminated by a collimated beam, rays at every height arrive parallel to the axis; the angle of incidence at a given radial zone is determined entirely by the surface slope at that height. When the IOL is illuminated by the converging, spherically-aberrated beam produced by the cornea, each ray arrives already inclined by an angle that varies with pupil radius. The total angle of incidence at each zone is the vector sum of the surface slope and this input ray angle. Because Snell’s law is nonlinear, this sum produces a different refracted direction than either contribution alone would suggest.

The critical consequence is this: two IOL surface geometries that produce the same wavefront output from a collimated input are, in general, not degenerate with respect to a curved or aberrated input. Their surfaces have different local curvatures and slopes at every zone; these differences are inconsequential when all input rays are axial but become consequential when input rays carry radially-varying inclinations. The simulation demonstrates this exactly: under collimated illumination, Configs. 1 and 2 are indistinguishable (Z11 = 0.327 waves, RMS ≈ 1.24 waves in both); under corneal illumination, Z11 is 0.174 waves in Config. 1 and 0.493 waves in Config. 2 – a factor of 2.8 difference.

Physical summary: The cornea does not change which rays reach the IOL (the stop fixes that), but it changes the direction of each ray when it arrives. The IOL’s surface geometry then refracts these obliquely-incident rays in a way that depends on the specific surface shape – a dependence that is invisible to a collimated-beam wavefront measurement. Two lenses that are ‘the same’ under collimated light are not the same optical systems under converging light.

Formally, the system wavefront can be written as a functional of both the IOL surface geometry G and the input wavefront WIn:

Wsystem = Φ(G, WIn)

where Φ is a nonlinear operator embodying Snell’s law at each surface. The virtual-addition approximation replaces this with:

Wsystem≈ W₁(Wflat) + W₂(Wflat)

This is valid only when Φ is linear in WIn, i.e., in the paraxial limit. Outside the paraxial regime, the correct expression is:

Wsystem = W₁(Wflat) + W₂(W₁)

and the second term – the IOL’s response to the corneal wavefront – cannot be obtained from any measurement of the IOL under collimated illumination alone.

7.2 Implications for ISO 11979-2 compliance

ISO 11979-2 requires a physical model cornea in the measurement path, not as an arbitrary convention, but because the standard’s framers recognized that the IOL’s optical performance is conditioned on the input wavefront it receives. A measurement protocol that evaluates the IOL under collimated illumination and then adds a nominal corneal wavefront in post-processing commits a physically unjustified linearization. The result is a characterization that:

(1) Cannot distinguish between IOL geometries that behave identically under collimated illumination but differently under corneal illumination, as demonstrated here;

(2) Systematically misrepresents the spherical aberration balancing between cornea and IOL;

(3) Produces MTF predictions that may differ substantially from those obtained with the cornea physically present, undermining the standard’s intent.

7.3 Conditions under which the approximation improves

The wavefront-addition approximation improves as lens surfaces become weaker (thin-lens limit), as the pupil diameter decreases (paraxial limit), and as the IOL refractive index approaches that of the surrounding medium. For modern IOLs with significant surface curvature, asphericity, and an entrance pupil diameter of 3 mm or above, the error is unlikely to be negligible relative to ISO 11979-2 tolerances. The present simulation, using a 5.15 mm diameter entrance pupil, produces a Z11 difference of 0.32 waves between two nominally equivalent designs – a difference that would produce a clearly measurable MTF discrepancy at spatial frequencies relevant to visual function.

8. CONCLUSION

Two intraocular lens designs engineered to produce identical wavefront errors under collimated illumination exhibit substantially different optical performance once a model cornea is placed anterior to them in a ray-trace simulation. The aperture stop at the IOL ensures that ray heights at the lens are identical in both measurement conditions; the discrepancy arises entirely because the cornea changes the direction of every ray incident on the IOL. Each IOL surface refracts these directionally-structured rays according to its specific local geometry via the nonlinear Snell’s law, producing a surface-shape-dependent aberration contribution that is invisible to any measurement performed under collimated illumination. Primary spherical aberration (Z11) differs by a factor of 2.8 and pupil-integrated power variation differs by a factor of 2.5 between the two designs once the corneal beam is present. A virtual corneal wavefront added numerically to a collimated-beam IOL measured wavefront cannot reproduce this angle-of-incidence-dependent refraction. ISO 11979-2’s requirement for a physical model cornea is therefore not merely a historical convention but a necessary condition for measuring IOL MTF in a manner that is representative of in-eye performance.