Fengyu Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fengyu Zhang has a total of 92 co-authors. Here are the top co-authors by number of shared publications:
- Antoine Kahn
Department of Electrical and Computer Engineering Princeton University Princeton NJ 08544 USA
11 shared publications - Barry P. Rand
Princeton University, USA
6 shared publications - Stephen Barlow
School of Chemistry and Biochemistry and Center for Organic Photonics and Electronics, Georgia Institute of Technology, Atlanta, GA 30332-0400, USA
3 shared publications - Birjodh Tiwana
LinkedIn, Mountain View, CA, USA
2 shared publications - Daqi Sun
LinkedIn, Mountain View, CA, USA
2 shared publications - Henry James Snaith
Clarendon Laboratory, Department of Physics, University of Oxford, Oxford, UK
2 shared publications
Latest Publications
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 6038-6044
LinkedIn Post Embeddings: Industrial Scale Embedding Generation and Usage across LinkedIn
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 6038-6044
Open AccessLinkedIn Post Embeddings: Industrial Scale Embedding Generation and Usage across LinkedIn
- Open Access
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 4804-4815
LiRank: Industrial Large Scale Ranking Models at LinkedIn
We present LiRank, a large-scale ranking framework at LinkedIn that brings to production state-of-the-art modeling architectures and optimization methods. We unveil several modeling improvements, including Residual DCN, which adds attention and residual connections to the famous DCNv2 architecture. We share insights into combining and tuning SOTA architectures to create a unified model, including Dense Gating, Transformers and Residual DCN. We also propose novel techniques for calibration and describe how we productionalized deep learning based explore/exploit methods. To enable effective, production-grade serving of large ranking models, we detail how to train and compress models using quantization and vocabulary compression. We provide details about the deployment setup for large-scale use cases of Feed ranking, Jobs Recommendations, and Ads click-through rate (CTR) prediction. We summarize our learnings from various A/B tests by elucidating the most effective technical approaches. These ideas have contributed to relative metrics improvements across the board at LinkedIn: +0.5% member sessions in the Feed, +1.76% qualified job applications for Jobs search and recommendations, and +4.3% for Ads CTR. We hope this work can provide practical insights and solutions for practitioners interested in leveraging large-scale deep ranking systems.
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 4804-4815
Open AccessLiRank: Industrial Large Scale Ranking Models at LinkedIn
We present LiRank, a large-scale ranking framework at LinkedIn that brings to production state-of-the-art modeling architectures and optimization methods. We unveil several modeling improvements, including Residual DCN, which adds attention and residual connections to the famous DCNv2 architecture. We share insights into combining and tuning SOTA architectures to create a unified model, including Dense Gating, Transformers and Residual DCN. We also propose novel techniques for calibration and describe how we productionalized deep learning based explore/exploit methods. To enable effective, production-grade serving of large ranking models, we detail how to train and compress models using quantization and vocabulary compression. We provide details about the deployment setup for large-scale use cases of Feed ranking, Jobs Recommendations, and Ads click-through rate (CTR) prediction. We summarize our learnings from various A/B tests by elucidating the most effective technical approaches. These ideas have contributed to relative metrics improvements across the board at LinkedIn: +0.5% member sessions in the Feed, +1.76% qualified job applications for Jobs search and recommendations, and +4.3% for Ads CTR. We hope this work can provide practical insights and solutions for practitioners interested in leveraging large-scale deep ranking systems.
- 13 December 2021
- Published by American Chemical Society (ACS) in ACS Energy Letters
- Vol. 7 (1) , 211-216
Electrochemically n-Doped CsPbBr3 Nanocrystal Thin Films
- 13 December 2021
- Published by American Chemical Society (ACS) in ACS Energy Letters
- Vol. 7 (1) , 211-216
Electrochemically n-Doped CsPbBr3 Nanocrystal Thin Films
- 1 October 2021
- Published by AIP Publishing in Applied Physics Reviews
- Vol. 8 (4) , 041301
Molecular dopants: Tools to control the electronic structure of metal halide perovskite interfaces
In the standard configurations of metal halide perovskite solar cell, the active layer, or absorber, follows a p-i-n or n-i-p electronic structure that is designed to enhance the separation and extraction of photo-induced charge carriers. The control of the Fermi level position across the film, between electron and hole transport layers, is therefore of paramount importance. Direct localized doping in metal halide perovskites being still elusive, the design of n-i-p and p-i-n structures has so far relied predominantly on surface and interface doping of the perovskite as well as on the control of the work function of the substrate and transport layers on which, or between which, the absorber is being placed. We provide here a short review of that work, emphasizing the fundamental studies of electronic structure performed on systems modified with organic molecular dopants. The review starts with a justification for the effectiveness of interface doping, based on the ability to move the Fermi level across the gap of the perovskite. We then review work done on the deposition of molecular oxidants and reductants on perovskite surfaces, including the mitigation of the surface states, and the impact of these dopants on energy level alignment with substrate and charge transport layers. The second part of the review focuses on the use of molecular dopants to either modify the work function of electron or hole transport layers to establish the boundary conditions for a p-i-n or n-i-p structure, or to enhance the conductivity of these layers in order to facilitate charge carrier extraction. Final considerations are also given on recent work on bulk doping of the perovskite layer with molecular dopants.
- 1 October 2021
- Published by AIP Publishing in Applied Physics Reviews
- Vol. 8 (4) , 041301
Molecular dopants: Tools to control the electronic structure of metal halide perovskite interfaces
In the standard configurations of metal halide perovskite solar cell, the active layer, or absorber, follows a p-i-n or n-i-p electronic structure that is designed to enhance the separation and extraction of photo-induced charge carriers. The control of the Fermi level position across the film, between electron and hole transport layers, is therefore of paramount importance. Direct localized doping in metal halide perovskites being still elusive, the design of n-i-p and p-i-n structures has so far relied predominantly on surface and interface doping of the perovskite as well as on the control of the work function of the substrate and transport layers on which, or between which, the absorber is being placed. We provide here a short review of that work, emphasizing the fundamental studies of electronic structure performed on systems modified with organic molecular dopants. The review starts with a justification for the effectiveness of interface doping, based on the ability to move the Fermi level across the gap of the perovskite. We then review work done on the deposition of molecular oxidants and reductants on perovskite surfaces, including the mitigation of the surface states, and the impact of these dopants on energy level alignment with substrate and charge transport layers. The second part of the review focuses on the use of molecular dopants to either modify the work function of electron or hole transport layers to establish the boundary conditions for a p-i-n or n-i-p structure, or to enhance the conductivity of these layers in order to facilitate charge carrier extraction. Final considerations are also given on recent work on bulk doping of the perovskite layer with molecular dopants.
- 22 April 2021
- Published by Springer Nature in Nature Materials
- Vol. 20 (9) , 1248-1254
Adduct-based p-doping of organic semiconductors
Electronic doping of organic semiconductors is essential for their usage in highly efficient optoelectronic devices. Although molecular and metal complex-based dopants have already enabled significant progress of devices based on organic semiconductors, there remains a need for clean, efficient and low-cost dopants if a widespread transition towards larger-area organic electronic devices is to occur. Here we report dimethyl sulfoxide adducts as p-dopants that fulfil these conditions for a range of organic semiconductors. These adduct-based dopants are compatible with both solution and vapour-phase processing. We explore the doping mechanism and use the knowledge we gain to ‘decouple’ the dopants from the choice of counterion. We demonstrate that asymmetric p-doping is possible using solution processing routes, and demonstrate its use in metal halide perovskite solar cells, organic thin-film transistors and organic light-emitting diodes, which showcases the versatility of this doping approach.
- 22 April 2021
- Published by Springer Nature in Nature Materials
- Vol. 20 (9) , 1248-1254
Adduct-based p-doping of organic semiconductors
Electronic doping of organic semiconductors is essential for their usage in highly efficient optoelectronic devices. Although molecular and metal complex-based dopants have already enabled significant progress of devices based on organic semiconductors, there remains a need for clean, efficient and low-cost dopants if a widespread transition towards larger-area organic electronic devices is to occur. Here we report dimethyl sulfoxide adducts as p-dopants that fulfil these conditions for a range of organic semiconductors. These adduct-based dopants are compatible with both solution and vapour-phase processing. We explore the doping mechanism and use the knowledge we gain to ‘decouple’ the dopants from the choice of counterion. We demonstrate that asymmetric p-doping is possible using solution processing routes, and demonstrate its use in metal halide perovskite solar cells, organic thin-film transistors and organic light-emitting diodes, which showcases the versatility of this doping approach.
- 3 September 2020
- Published by Wiley in Advanced Materials
- Vol. 32 (42) , e2003482
Gap States in Methylammonium Lead Halides: The Link to Dimethylsulfoxide?
Understanding the origin and distribution of electronic gap states in metal halide perovskite (MHP) thin films is crucial to the further improvement of the efficiency and long‐term stability of MHP‐based optoelectronic devices. In this work, the impact of Lewis‐basic additives introduced in the precursor solution on the density of states in the perovskite bandgap is investigated. Ultraviolet photoemission spectroscopy and contact potential difference measurements are conducted on MHP thin films processed from dimethylformamide (DMF)‐based solutions to which either no additive, dimethylsulfoxide (DMSO), or N‐methylpyrrolidine‐2‐thione (NMPT) is added. The results show the presence of a density of states in the gap of methylammonium lead halide films processed from DMSO‐containing solution. The density of gap states is either suppressed when the methylammonium concentration in mixed cation films is reduced or when NMPT is used as an additive, and eliminated when methylammonium (MA) is replaced with cesium or formamidinium (FA). These results are consistent with the notion that reaction products that result from DMSO reacting with MA+ in the precursor solution are responsible for the formation of gap states.
- 3 September 2020
- Published by Wiley in Advanced Materials
- Vol. 32 (42) , e2003482
Gap States in Methylammonium Lead Halides: The Link to Dimethylsulfoxide?
Understanding the origin and distribution of electronic gap states in metal halide perovskite (MHP) thin films is crucial to the further improvement of the efficiency and long‐term stability of MHP‐based optoelectronic devices. In this work, the impact of Lewis‐basic additives introduced in the precursor solution on the density of states in the perovskite bandgap is investigated. Ultraviolet photoemission spectroscopy and contact potential difference measurements are conducted on MHP thin films processed from dimethylformamide (DMF)‐based solutions to which either no additive, dimethylsulfoxide (DMSO), or N‐methylpyrrolidine‐2‐thione (NMPT) is added. The results show the presence of a density of states in the gap of methylammonium lead halide films processed from DMSO‐containing solution. The density of gap states is either suppressed when the methylammonium concentration in mixed cation films is reduced or when NMPT is used as an additive, and eliminated when methylammonium (MA) is replaced with cesium or formamidinium (FA). These results are consistent with the notion that reaction products that result from DMSO reacting with MA+ in the precursor solution are responsible for the formation of gap states.
- 29 January 2020
- Published by Wiley in Advanced Energy Materials
- Vol. 10 (26) , 1903252
Ultraviolet Photoemission Spectroscopy and Kelvin Probe Measurements on Metal Halide Perovskites: Advantages and Pitfalls
In this essay, a case study is presented on the electronic structure of several metal halide perovskites (MHP) using Kelvin probe (KP)‐based surface photovoltage (SPV) measurements and ultraviolet photoemission spectroscopy (UPS) to demonstrate the advantages, but also the pitfalls, of using these techniques to characterize the surfaces of these materials. The first part addresses the loss of halide species from perovskite surfaces upon supragap illumination in vacuum. This has the potential to cause both a long‐term alteration of the sample work function and a modification of the KP tip during SPV measurements. If undetected, this leads to a misinterpretation of the MHP surface potential. The second part illustrates the difficulties in determining the valence band maximum (VBM) of MHP surfaces with UPS and stresses the importance of taking into account the low density of states at the VBM edge. Given this circumstance, specific care must be taken to eliminate measurement artifacts in order to ascertain the presence or absence of low densities of electronic gap states above the VBM. This essay also highlights issues such as film degradation, nonequilibrium situations (e.g., SPV), and satellite emissions, which occur during photoemission spectroscopy.
- 29 January 2020
- Published by Wiley in Advanced Energy Materials
- Vol. 10 (26) , 1903252
Ultraviolet Photoemission Spectroscopy and Kelvin Probe Measurements on Metal Halide Perovskites: Advantages and Pitfalls
In this essay, a case study is presented on the electronic structure of several metal halide perovskites (MHP) using Kelvin probe (KP)‐based surface photovoltage (SPV) measurements and ultraviolet photoemission spectroscopy (UPS) to demonstrate the advantages, but also the pitfalls, of using these techniques to characterize the surfaces of these materials. The first part addresses the loss of halide species from perovskite surfaces upon supragap illumination in vacuum. This has the potential to cause both a long‐term alteration of the sample work function and a modification of the KP tip during SPV measurements. If undetected, this leads to a misinterpretation of the MHP surface potential. The second part illustrates the difficulties in determining the valence band maximum (VBM) of MHP surfaces with UPS and stresses the importance of taking into account the low density of states at the VBM edge. Given this circumstance, specific care must be taken to eliminate measurement artifacts in order to ascertain the presence or absence of low densities of electronic gap states above the VBM. This essay also highlights issues such as film degradation, nonequilibrium situations (e.g., SPV), and satellite emissions, which occur during photoemission spectroscopy.
- 5 August 2019
- Published by Royal Society of Chemistry (RSC) in Energy & Environmental Science
- Vol. 12 (10) , 3063-3073
Interfacial charge-transfer doping of metal halide perovskites for high performance photovoltaics
The remarkable optoelectronic properties of metal halide perovskites have generated intense research interest over the last few years. The ability to control and manipulate the crystallisation and stoichiometry of perovskite thin-films has allowed for impressive strides in the development of highly efficient perovskite solar cells. However, being able to effectively modify the interfaces of metal halide perovskites, and to controllably p- or n-type dope the surfaces, may be key to further improvements in the efficiency and long-term stability of these devices. In this study, we use surface doping of the mixed-cation, mixed-halide perovskite FA0.85MA0.15Pb(I0.85Br0.15)3 (FA – formamidinium; MA – methylammonium) to improve the hole extraction from the perovskite solar cell. By treating the surface of the perovskite film with a strongly oxidizing molybdenum tris(dithiolene) complex, we achieve a shift in the work function that is indicative of p-doping, and a twofold increase in the total conductivity throughout the film. We probe the associated interfacial chemistry through photoelectron and solid-state nuclear magnetic resonance spectroscopies and confirm that charge-transfer occurs between the perovskite and dopant complex. The resulting p-doped interface constitutes a homojunction with increased hole-selectivity. With charge-selective layers, we show that this surface doping enhances the device performance of perovskite solar cells resulting in steady-state efficiencies approaching 21%. Finally, we demonstrate that a surface treatment with this dopant produces the same effect as the commonly employed additive 4-tert butylpyridine (tBP), allowing us to achieve “tBP-free” devices with steady-state efficiencies of over 20%, and enhanced thermal stability as compared to devices processed using tBP. Our findings therefore demonstrate that molecular doping is a feasible route to tune and control the surface properties of metal halide perovskites.
- 5 August 2019
- Published by Royal Society of Chemistry (RSC) in Energy & Environmental Science
- Vol. 12 (10) , 3063-3073
Interfacial charge-transfer doping of metal halide perovskites for high performance photovoltaics
The remarkable optoelectronic properties of metal halide perovskites have generated intense research interest over the last few years. The ability to control and manipulate the crystallisation and stoichiometry of perovskite thin-films has allowed for impressive strides in the development of highly efficient perovskite solar cells. However, being able to effectively modify the interfaces of metal halide perovskites, and to controllably p- or n-type dope the surfaces, may be key to further improvements in the efficiency and long-term stability of these devices. In this study, we use surface doping of the mixed-cation, mixed-halide perovskite FA0.85MA0.15Pb(I0.85Br0.15)3 (FA – formamidinium; MA – methylammonium) to improve the hole extraction from the perovskite solar cell. By treating the surface of the perovskite film with a strongly oxidizing molybdenum tris(dithiolene) complex, we achieve a shift in the work function that is indicative of p-doping, and a twofold increase in the total conductivity throughout the film. We probe the associated interfacial chemistry through photoelectron and solid-state nuclear magnetic resonance spectroscopies and confirm that charge-transfer occurs between the perovskite and dopant complex. The resulting p-doped interface constitutes a homojunction with increased hole-selectivity. With charge-selective layers, we show that this surface doping enhances the device performance of perovskite solar cells resulting in steady-state efficiencies approaching 21%. Finally, we demonstrate that a surface treatment with this dopant produces the same effect as the commonly employed additive 4-tert butylpyridine (tBP), allowing us to achieve “tBP-free” devices with steady-state efficiencies of over 20%, and enhanced thermal stability as compared to devices processed using tBP. Our findings therefore demonstrate that molecular doping is a feasible route to tune and control the surface properties of metal halide perovskites.
- 14 March 2019
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 31 (17) , 6624-6632
Molecular-Reductant-Induced Control of a Graphene–Organic Interface for Electron Injection
- 14 March 2019
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 31 (17) , 6624-6632
Molecular-Reductant-Induced Control of a Graphene–Organic Interface for Electron Injection
- 10 February 2019
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry Letters
- Vol. 10 (4) , 890-896
Complexities of Contact Potential Difference Measurements on Metal Halide Perovskite Surfaces
Understanding the stability of metal halide perovskite (MHP) surfaces is of considerable interest for the development of devices based on these materials. We present here a vacuum-based study of the surface potential and response to illumination of two different types of perovskite films, methylammonium lead bromide (MAPbBr3) and the 2D Ruddlesden–Popper phase butylammonium lead iodide (BA2PbI4, n = 1), using Kelvin probe-based contact potential difference and surface photovoltage measurements. We show that supraband gap light irradiation can induce the loss of halide species, which adsorb on the Kelvin probe tip, inducing quasi-irreversible changes of the MHP surface and tip work functions. If undetected, this can lead to misinterpretations of the MHP surface potential. Our results illustrate the effectiveness of the Kelvin probe-based technique in providing complementary information on the energetics of perovskite surfaces and the necessity to monitor the work function of the probe to avoid erroneous conclusions when working on these materials.
- 10 February 2019
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry Letters
- Vol. 10 (4) , 890-896
Complexities of Contact Potential Difference Measurements on Metal Halide Perovskite Surfaces
Understanding the stability of metal halide perovskite (MHP) surfaces is of considerable interest for the development of devices based on these materials. We present here a vacuum-based study of the surface potential and response to illumination of two different types of perovskite films, methylammonium lead bromide (MAPbBr3) and the 2D Ruddlesden–Popper phase butylammonium lead iodide (BA2PbI4, n = 1), using Kelvin probe-based contact potential difference and surface photovoltage measurements. We show that supraband gap light irradiation can induce the loss of halide species, which adsorb on the Kelvin probe tip, inducing quasi-irreversible changes of the MHP surface and tip work functions. If undetected, this can lead to misinterpretations of the MHP surface potential. Our results illustrate the effectiveness of the Kelvin probe-based technique in providing complementary information on the energetics of perovskite surfaces and the necessity to monitor the work function of the probe to avoid erroneous conclusions when working on these materials.