This paper aims to realize an ultra-low Narcissus design for a cooled MWIR refraction-diffraction hybrid optical system containing a Diffractive Optical Element （DOE）， while maintaining high imaging performance under a large relative aperture. The study is focused on three tasks： examining the applicability of conventional Narcissus evaluation criteria in a hybrid optical system， establishing a quantitative analysis route based on actual ray tracing， and obtaining an optical design with significantly suppressed Narcissus irradiance at the focal plane. The objective is to reduce residual Narcissus noise beyond the level achievable by routine low-Narcissus design， so that both image uniformity and temperature-measurement accuracy can be improved in a high-resolution cooled infrared imaging system.A combined qualitative and quantitative procedure was adopted. First， the conventional

YNI

and

I/IBAR

criteria， together with the diffractive correction of

YNI

for the DOE surface， were used to obtain an initial low-Narcissus solution and to evaluate the Narcissus tendency of individual optical surfaces. Second， in view of the limitations of these surface-based criteria in a comp

lex hybrid system， a quantitative evaluation framework based on NITD （Narcissus Induced Equivalent Temperature Difference） and reverse ray tracing was established to characterize the Narcissus effect at the focal plane. In this framework， the propagation paths of Narcissus rays were traced surface by surface， and the contribution of each optical surface to the focal-plane Narcissus irradiance was quantified. For the DOE， the analysis considered not only the design diffraction order， but also non-high-diffraction-efficiency wavelengths in the design order and the contribution of non-design diffraction orders. For the refractive part， the dominant Narcissus sources were identified through NITD contribution analysis and ray-path tracing. Based on these results， targeted optimization was carried out by adjusting the DOE structural parameters， modifying the optical power distribution of the rear group， and introducing additional aspheric degrees of freedom where necessary. The final design was implemented in ZEMAX for a 3–5 μm waveband， with a 165 mm focal length， F/1.65， 6° full field of view， 1024×1024 detector format， and 12 μm pixel size.The results show that YNI and I/IBAR can indicate the Narcissus tendency of individual surfaces， but they cannot accurately represent the actual Narcissus level of a complex refraction-diffraction hybrid optical system. Even when the calculated

YNI

or

I/IBAR

values of some surfaces appear favorable， the corresponding surfaces may still produce substantial contributions to the total NITD after full-path ray tracing is considered. For the DOE， the analysis demonstrates that Narcissus is not determined only by the nominal design-order rays； rays at non-high-efficiency wavelengths in the design order and rays generated by non-design diffraction orders also contribute to the focal-plane Narcissus irradiance. In the initial design， the maximum detector NITD reached approximately 0.2309 K， and the dominant contribution originated from the rear group， wit

h surface S7 acting as the principal Narcissus source. After targeted optimization， the irradiance of DOE-related Narcissus rays arriving at the focal plane decreased from about 270.4 to 160.9， corresponding to a reduction of approximately 40.5%. At the same time， the maximum total NITD was reduced to 0.0490 K. The dominant Narcissus contribution shifted from the original surface to S10， while the contributions of most other surfaces became lower and more balanced. In the optimized system， the Narcissus associated with S7， S8， and S9 was effectively suppressed， and the rear group was reconfigured by using aspheric surfaces at S7， S8， and S13， together with increased curvature of S9 and S10， so that reflected Narcissus rays were less likely to be recollected onto the focal plane. Meanwhile， the system maintained good imaging performance over the operating temperature range.For a cooled MWIR refraction-diffraction hybrid optical system， ultra-low Narcissus design cannot depend solely on conventional surface-based criteria such as

YNI

and

I/IBA

R. A more effective design route is to combine preliminary criterion-based screening with NITD-oriented actual ray tracing， then perform path-oriented optimization aimed at both suppressing Narcissus generation and weakening the recollection of Narcissus rays by the rear group. When the original structural degrees of freedom are insufficient， additional optical-surface freedom， especially the introduction of aspheric surfaces， is necessary to balance imaging performance and Narcissus suppression. The final design verifies that an ultra-low Narcissus level can be achieved in a DOE-containing cooled MWIR imaging system， and indicates that the key issue in further improvement lies in controlling the localized distribution of residual Narcissus irradiance on the focal plane.