Junzhi Ye 1,2#, Navendu Mondal 3#*, Ben P. Carwithen 3, Yunwei Zhang 4, Linjie Dai 1,5, Xiang-Bing Fan 6, Jian Mao5,7, Zhiqiang Cui 4, Pratyush Ghosh 1, Clara Otero‐Martínez 8, Lars van Turnhout1, Yi-Teng Huang2, Zhongzheng Yu 1, Ziming Chen 3, Neil C. Greenham 1, Samuel D. Stranks 1, 5, Lakshminarayana Polavarapu 8, Artem Bakulin 3, Akshay Rao 1, Robert L.Z. Hoye 2,9*
1.	Cavendish Laboratory, University of Cambridge, 11880, Cambridge CB3 0HE, United Kingdom
2.	Inorganic Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QR, United Kingdom
3.	Department of Chemistry and Centre for Processable Electronics, Imperial College London, Molecular Sciences Research Hub, 83 Wood Lane, London W12 0BZ, United Kingdom 
4.	School of Physics, Sun Yat-sen University, 510275 Guangzhou, China
5.	Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge CB3 0AS, United Kingdom. 
6.	Department of Engineering, University of Cambridge, 9 JJ Thomson Avenue, Cambridge, CB3 0FA, United Kingdom
7.	State Key Laboratory of Photovoltaic Science and Technology, Shanghai Frontiers Science Research Base of Intelligent Optoelectronics and Perception, Institute of Optoelectronics, Fudan University, Shanghai, 200433, China
8.	CINBIO, Universidade de Vigo, Materials Chemistry and Physics Group, Department of Physical Chemistry, Campus Universitario As Lagoas, Marcosende, 36310 Vigo, Spain
9.	Department of Materials, Imperial College London, Exhibition Road, London SW7 2AZ, United Kingdom. 
# These authors contributed equally to this work.
* Email: n.mondal@imperial.ac.uk (N.M.),  robert.hoye@chem.ox.ac.uk (R. L. Z. H.)


Main Text: 
Raw data is store in txt or origin files. It can be opened in origin, adobe illustrator, matlab or excel. 

Fig.1
a: Change in the normalized photoluminescence quantum yield (PLQY) of CsPbI3, CsPbBrxI3-x and CsPbBr3 nanocrystals after successive purification steps. All PLQY values were normalized at the lowest excess energy. Comparison of the photoluminescence (PL) and absorption spectra of pristine (low defect density), singly purified (moderate defect density), and doubly purified (high defect density). Data obtained by Junzhi Ye with liquid samples in quartz cuvette using a 400 nm CW laser and an integrating sphere.
b-d: Nanocrystal solution PL and UV-Vis Absorption Spectra obtained in 1 mm quartz cuvette. Data obtained by Junzhi Ye.
e-g: Excitation-wavelength-dependent PLQY for CsPbBr3, CsPbBrxI3-x, and CsPbI3 nanocrystals solutions in quartz cuvette. Data obtained by Junzhi Ye and Xiangbin Fan.

Fig.2
a-b: Transient absorption spectra for low and high defect Br-nancrystal solution measured in quartz cuvette using 400 nm pulse laser (500 Hz rep rate and 116 uJ cm-2 fluence). Data obtained by Junzhi Ye
c-e: Transient absoprtion kinetics for the sub-bandgap states for br, B/I and I-based NCs with decovolution of photo-induced absoprtion and trap-bleach. The data is obtained by Junzhi Ye, and the SVD is performed by Yi-Teng Huang. 

Fig.3
a-i: Hot carrier energy loss rate based on pump-probe transient absorption spectroscopy measurements. TA maps for colloidal solutions of low, moderate and high defect density a-c, CsPbBr3, d-f, CsPbBrxI3-x and g-i, CsPbI3 perovskite nanocrystals under 400 nm wavelength excitation. Data obtained by Junzhi Ye.
j-l: Energy loss rate for different defect concentrations in j, CsPbBr3, k, CsPbBrxI3-x, and l, CsPbI3 NCs. Data obtained by Junzhi Ye.

Fig.4 
a-d: HC lifetime (cool) obtained by fitting from pump-probe transient absorption spectroscopy measurements. Data obtained by Junzhi Ye.

Fig.5:
b: Exemplar comparison of the pump-probe (PP) and pump-push-probe (PPP) GSB decay kinetics with a pump-push delay of 10 ps for low defect density CsPbBr3 NC solution.
c: Hot carrier density (or push-fluence) dependent representative GSB decay curves. 
d-f: HC lifetimes obtained from fitting the PPP TA measurements for d, CsPbBr3, e, CsPbBrxI3-x, and f, CsPbI3 NCs, respectively. 
Data obtained by Navendu Mondal and Ben P. Carwithen.

Supplementary Information:

Supplementary Fig.2
a-c: a, Relative Pb/Br ratio for CsPbBr3 NCs after washing. b. Relative Pb/Br and I ratio for CsPbBrxI3-x NCs after washing. c. Relative Pb/I ratio for CsPbI3 NCs after washing. The ratio is caluculated based on the integrated area for halide 3d and Pb 4f core level spectra shown in Supplementary Fig.3. XPS data obtained by Dr. Mark Issac from Harwell XPS.
d-f: Photothermal deflection spectroscopy (PDS) measurement for CsPbX3 NCs with different defect densities and their Urbach energy. The PDS results and Urbach energy fitting is measured and calculated following our previous work (https://doi.org/10.1021/jacs.2c02631). Data obtained by Lars van Turnhout. 


Supplementary Fig.3
a-f: Compositional analysis of X-ray photoemission spectroscopy. a-c, Halide (I and Br) 3d core level XPS spectra. d-f, Pb 4f core level XPS spectra of low, moderate and high defect samples. The integrated area ratio between halide and Pb are shown in Supplementary Fig. 2. XPS data obtained by Dr. Mark Issac from Harwell XPS.

Supplementary Fig.4
Time-correlated Single Photon Counting Measurements. Influence of defects densities on photoluminescence lifetime for a, CsPbBr3, b, CsPbBrxI3-x and c, CsPbI3 NCs solution in quartz cuvette. Measured with a 405 nm wavelength pulsed laser. Data obtained by Linjie Dai and Jian Mao.


Supplementary Fig.6
a-f: Supplementary Fig. 6 | Computational investigation into the effect of defect states in the bulk of CsPbX3 on their electronic structure and density of states. Band structures and projected density of states of pristine (no defects) a, CsPbBr3, b, CsPbBrxI3-x, and c, CsPbI3, along with defective d, CsPbBr3, e, CsPbBrxI3-x, and f, CsPbI3. The defective materials were simulated as having two halide vacancies per unit cells. Please note that this simulation was performed on bulk materials, not NCs, and does not account for surface defects. Details of how these calculations were performed are provided in the Methods section. The calculations are performed by Zhiqiang Cui and Yunwei Zhang. 

Supplementary Fig.7
a-d: Short-time transient absorption signal decomposition. Spectra deconvolution for a, low defects CsPbBrxI3-x NCs and b, high defects CsPbBrxI3-x NCs. Spectra deconvolution for c, low defects CsPbI3 NCs and d, high defects CsPbI3 NCs. The decomposition method is used in our previous report (https://doi.org/10.1038/s41467-022-32669-3). SVD is done by Yi-Teng Huang. 

Supplementary Fig.8
a-c: TA maps of CsPbBr3 NCs with defect densities from low to high. 
d-f:TA spectra of CsPbBr3 NCs with defect densities from low to high. 
h-g: Fitted hot carrier cooling lifetime with different fluence for CsPbBr3 NCs. 
The pump wavelength was 400 nm, repetition rate 500 Hz, and the maps and spectra were recorded under a fluence of 194.32 μJ cm-2 using NC solution in quartz cuvette.
Data obtained by Junzhi Ye. 

Supplementary Fig.9
a-c: TA maps of CsPbBrI3 NCs with defect densities from low to high. 
d-f:TA spectra of CsPbBrI3 NCs with defect densities from low to high. 
h-g: Fitted hot carrier cooling lifetime with different fluence for CsPbBrI3 NCs. 
The pump wavelength was 400 nm, repetition rate 500 Hz, and the maps and spectra were recorded under a fluence of 237.58 μJ cm-2 using NC solution in quartz cuvette.
Data obtained by Junzhi Ye. 

Supplementary Fig.10
a-c: TA maps of CsPbI3 NCs with defect densities from low to high. 
d-f:TA spectra of CsPbI3 NCs with defect densities from low to high. 
h-g: Fitted hot carrier cooling lifetime with different fluence for CsPbI3 NCs. 
The pump wavelength was 400 nm, repetition rate 500 Hz, and the maps and spectra were recorded under a fluence of 193.36 μJ cm-2 using NC solution in quartz cuvette.
Data obtained by Junzhi Ye. 


Supplementary Fig.11
Selection of the push energy. To select the appropriate wavelength for the push pulse, the IR transient absorption was performed in the range between 900 nm to 1450 nm wavelength. The highest absorption was reached at 1300 nm wavelength, hence its selection as the push beam wavelength for the PPP analysis. Data obtained by Navendu Mondal and Ben P. Carwithen.

Supplementary Fig.12
Kinetics of GSB and PAs for the representative case of CsPb(Br/I)3 NCs under PP and PPP-TA measurements. All samples are colloidal solutions, and were measured inside a 1 mm thick cuvette. The NIR push beam has an energy of 0.95 eV (1300 nm). The pump laser is a 400 nm pulse laser at 2 μJ cm-2. Data obtained by Navendu Mondal and Ben P. Carwithen.