Journal Pre-proof

Life-cycle energy demand comparison of medium voltage Silicon

IGBT and Silicon Carbide MOSFET power semiconductor modules in
railway traction applications

Lucas Barroso Spejo , Innocent Akor , Munaf Rahimo ,

Renato Amaral Minamisawa

PII: S2772-3704(23)00018-4
DOI: https://doi.org/10.1016/j.pedc.2023.100050
Reference: PEDC 100050

To appear in: Power Electronic Devices and Components

Received date: 7 August 2023

Revised date: 4 November 2023
Accepted date: 5 November 2023

Please cite this article as: Lucas Barroso Spejo , Innocent Akor , Munaf Rahimo ,
Renato Amaral Minamisawa , Life-cycle energy demand comparison of medium voltage Silicon
IGBT and Silicon Carbide MOSFET power semiconductor modules in railway traction applications,
Power Electronic Devices and Components (2023), doi: https://doi.org/10.1016/j.pedc.2023.100050

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2023 Published by Elsevier Ltd.

This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
HIGHLIGHTS

• Use of specialized fab data for cumulative energy demand comparison of power modules.
• SiC has higher energy demand per area than Si, but lower power module grey energy.
• Power module manufacturing energy is negligible compared to the use-phase energy.
• Significant potential energy saving when using SiC modules in railway inverters.

1
Running title: Life-cycle comparison of Si IGBT and SiC MOSFET modules for railway traction
Power Electronic Devices and Components
Research Article

Life-cycle energy demand comparison of medium voltage Silicon IGBT and Silicon Carbide

MOSFET power semiconductor modules in railway traction applications

Lucas Barroso Spejo1, Innocent Akor1, Munaf Rahimo2, Renato Amaral Minamisawa1

1
Institute of Electric Power Systems, University of Applied Sciences and Arts Northwestern Switzerland, 5210
Windisch, Switzerland
2
MTAL GmbH, 4716 Gänsbrunnen, Switzerland

Abstract

Power semiconductors process roughly 70 % of global energy, with a higher percentage expected as worldwide
transport electrification, renewables and wide-band-gap (WBG) semiconductors are implemented, significantly
affecting global energy savings. This manuscript evaluates the cumulative energy demand (CED) encompassing
the manufacture and use-phase in a railway traction application of silicon (Si) and silicon carbide (SiC) power
semiconductor modules. Realistic manufacturing data from a power semiconductor fab has been considered for
3.3 kV/450 A state-of-the-art Si and SiC LinPak modules. SiC devices presented around 2.6 – 3.8× higher CED
per area than Si devices in the manufacturing phase. However, due to the considerably smaller SiC chip area per
ampere required, a 1.1 – 1.6× lower grey energy than Si technology is estimated. For the first time, such analysis
is based on specialized power semiconductor fab data for both technologies and provides a baseline for the life
cycle energy assessment of power electronics systems. Besides, the use-phase energy losses were evaluated for a
realistic railway application, considering an operational lifetime of 30 years. The module manufacturing energy
is negligible compared to the use-phase stage. Furthermore, the SiC technology presented an estimated energy-
saving potential of 24 MWh/lifetime per module compared to the Si device.

Keywords: Silicon carbide MOSFET, Silicon IGBT, Life cycle energy assessment, Power semiconductor
module, Railway traction, Power electronics

Lucas B. Spejo (corresponding author)

lucas.spejo@fhnw.ch
Renato A. Minamisawa (contact)
renato.minamisawa@fhnw.ch

1. Introduction

Power semiconductor devices are the key elements responsible for energy conditioning in power
converters. Some typical applications are wind and solar energy generation, electric vehicle traction, battery
charging, railway traction, industrial motor drives, energy transmission, data center power supplies, and others

2
[1]. Approximately 70 % of global energy is already processed by power devices [2], and a higher percentage
may be expected with the worldwide implementation of renewables, WBG semiconductors, and transport
electrification. Such motivation is caused by actual carbon dioxide reduction policies. In this sense, Railway
transport is one of the most efficient and environmentally friendly options with a high traffic density [3]. Such
characteristics make trains an excellent option for expanding transport systems and decreasing global carbon
dioxide emissions [4].

The mainstream technology of power semiconductors is still based on silicon due to its good
performance, high reliability and accessible cost. However, power switches based on wide band gap (WBG)
semiconductors (e.g. silicon carbide) can improve performance by lowering energy losses and reducing power
converter size and weight [1]. Such characteristics are possible because SiC presents a bandgap approximately
three times larger than Si, which allows devices with higher blocking voltages and lower ohmic resistance.
Additionally, it presents lower switching losses enabling higher operating frequencies and, consequently, smaller
power converters [1, 5]. The major drawback that still restricts the full adoption of the WBG technology is higher
costs than Si (around 3-4 times). Such fact is caused by the highly energy demanding manufacturing steps required
during material synthesis and device processing [6]. Nevertheless, several industrial products already present SiC
in their power converters. Fig. 1 shows the timeline of the main achievements from 3.3 kV SiC power modules in
railway technology. In the last 3 years, there was a successful implementation in train lines worldwide [7, 8].
Besides, a further commitment to expand its implementation with cooperation between Mitsubishi Electric and
Siemens Mobility was signed in 2022 [9]. With further development of the SiC technology, lower costs and higher
technology maturity may be achieved, transforming SiC into the mainstream technology in the future.

The energy demand assessment of semiconductor device technologies in power converters systems is
essential to investigate their energy impact and benefits. Life cycle assessment of systems that utilize power
electronics converters like electric vehicles [10-13], photovoltaic and wind energy systems [14-17] are widely
investigated in the literature. However, LCA researchers face a challenge: limited data available from the energy
spent on the manufacturing of power devices. Generally, LCA practitioners employ the energy consumed in
silicon device manufacturing processes from other applications in their models. For example, Ref [18] used
dynamic random-access memory (DRAM) manufacturing data [19] for Si Insulated-gate bipolar transistor (IGBT)
power switches due to the lack of publicly available life cycle data. The literature also presents data on the
manufacturing of silicon-based microelectronics devices and solar cells [20]. All these devices present distinct
front-end and back-end processing steps from typical power devices like Si Insulated-gate bipolar transistors
(IGBTs) and Metal-oxide-semiconductor-field-effect transistors (MOSFETs). Thus, accurate analysis of power
electronics systems becomes difficult due to the need for more specialized data, especially in the front-end stage,
where most of the manufacturing energy is consumed [18, 19, 21, 22]. Such an issue is mainly caused by the
complexity of semiconductor processing, industrial process confidentiality and rapid technology development
with short product lifetimes [23]. For example, a typical IGBT processing may require about 12 masks and 240
steps in the front-end processing, whereas DRAM presents around 400 steps and solar cells five to seven steps
[20]. Besides, other variables intrinsically related to the semiconductor plant type, such as the fab automation
level, fab size and plant utilization, play a key role in the final energy demand and add more uncertainty to
estimated values. Consequently, for better comparisons between different technologies (e. g. Si IGBT and SiC

3
MOSFET), the use of manufacturing data from the same fab, especially from the front-end stage, is essential.
Thus, the semiconductor plant characteristics will be the same, and uncertainties may be reduced.

Regarding life cycle analysis (LCA) analyses of silicon carbide devices in typical power electronic
applications, the literature is even more scarce due to the recent SiC technology market introduction. Ref
[18] performed a life cycle analysis of the demanded energy in electric cars using SiC MOSFET technology. An
industry consultant's teardown analysis of 600 V SiC Schottky diodes from several manufacturers has been used,
which may differ significantly from SiC MOSFETs power module processing parameters. In the work of [24],
the main differences between Si and SiC manufacturing were highlighted. Although, a deeper investigation could
not be performed due to the scarce available literature data with sufficient detail and quality. Our work provides
specialized front-end data from a power semiconductor fab for both technologies. The authors believe that the
analysis provided here contributes to the power electronics/semiconductors industry, academic researchers and
LCA practitioners to evaluate the manufacturing semiconductors processes efficiency as well as to provide a
baseline for the analysis of power electronics systems in general. It can also inform decision-makers about the
environmental implications of such technologies to help formulate governmental policies, including tax
incentives. It is essential to reinforce the scarcity of specialized data in the literature from power semiconductor
plants that are not readily available by manufacturers. Such data heavily depends on the fab type and imposes
accuracy limitations on the life cycle energy analysis of power systems.

The main scientific contributions of this paper are:

● To provide an energy assessment comparison of state-of-the-art LinPak 3.3 kV/450 A power modules
for Si and SiC technology. The analysis encompasses the most energy-demanding steps of the product life-cycle
and not only the end-use application, as it is widely reported in the literature [7, 25, 26].

● To report for the first time on realistic energy consumption data from a power semiconductor fab
manufacturing both technologies. It is also relevant to mention that the semiconductor manufacturing data
provided in this paper can serve as a reference for different applications that use 3.3 kV power modules. A few
examples are grid-tied converters [27], industrial drives [28], HVDC [29], and Solid-state transformers [30].
Furthermore, the data can also be used as a baseline for other voltage class gate-controlled devices (e. g. 1.2 kV,
1.7 kV, 4.5 kV and 6.5 kV Si IGBTs and SiC MOSFETs) with a reasonable accuracy. Since this data is not readily
available from power semiconductor manufacturers, the authors believe the data provided here allows future CED
analyses from several power electronics applications to have greater accuracy. Topics of huge interest due to their
global energy impacts, such as transport electrification, renewables and WBG semiconductors may be benefited.

● A detailed simulation based on an actual train application was also performed to compare both
modules and to depict the WBG technology potential in the use-phase, as well as to perform the lifetime CED
comparison. The 3.3 kV SiC MOSFET power module is a recent technology that already presents excellent energy
saving prospects compared to Si IGBT technology in railway applications [7, 8, 25, 26, 31]. This technology also
presents intensive development in recent years and a market commitment for further expansion (see Fig. 1).

The paper is organized as follows: Section 2 explains the methodology regarding data acquisition and
assumptions. Section 3 details the power module fabrication steps and their manufacturing cumulative energy

4
demand. Section 4 provides the use-phase cumulative energy demand per power module. Finally, section 5 is
dedicated to conclusions.

Fig. 1 Timeline of the main events from the 3.3 kV SiC power module technology in railway traction applications.

2. Methodology

2.1. Materials acquisition and manufacturing energy

Semiconductors devices present a highly complex manufacturing process that can be divided into three
main categories: (1) Substrate production, (2) Front-end, and (3) Back-end processing, as indicated in Fig. 2. The
first step comprises the processes performed from material extraction until wafer fabrication. The front-end stage
encompasses the physical and chemical processes performed in a fab environment to transform the wafer into
power device chips. Finally, the back-end processing corresponds to wafer die cutting and packaging. All steps
are critical from a technical point of view and essential for achieving a reliable functional product. However,
regarding cumulative energy demand (CED), the back-end processing has a minor contribution, being not the
focus of most investigations [18]. In addition, in our work, the Si and SiC LinPak packaging fabricated by Hitachi
Energy [27, 32] presents the same footprint, design and materials. Thus, the packaging production energy has not
been considered since it is roughly the same for both technologies and will not impact the final energy demand
comparison. The LinPak packaging comprises a protecting case with external electrical contacts and a base plate
for heat transfer (Fig. 3A). Internally, four substrates, as shown in Fig. 3B, are placed with bonded chips on the
copper surface. Semiconductor chips are electronic devices that operate as a switch controlled by an external
signal from a controller. These chips can operate in a blocking state (open switch), with no current flowing through
them, and in conducting state (closed switch), with the electrical current flowing. Such operation is responsible
for the energy conditioning in power converters. The chips are soldered on the substrate top copper layer (see Fig.
3B). These substrates are made of two copper layers electrically isolated by an alumina layer. Alumina is an
insulator material responsible for the electrical isolation of the bottom copper layer from the chips. This bottom
layer is soldered to the external base plate made of AlSiC. This plate is the thermal connection of the chips to the
external cooling system, and it is electrically neutral for safety reasons. The electrical chip connections are
distributed on the top copper layer through a wire bonding technique, and then the external electrical connections
are made to the users. The LinPak design is part of a class of standard state-of-the-art power modules based on a
joint effort of several manufacturers. The objective was to fabricate a more reliable power module generation with

5
improved interconnect technologies, low stray-inductance and compatibility with WBG chips [33]. These devices
are aimed to be used in medium voltage applications, with each manufacturer providing a different name for a
similar case design, as shown in Fig 4. Different current ratings of SiC LinPaks are presented at 450 A [27], 500
A [34] and 900 A [27].

Fig. 2 Power module life cycle. Substrate, front-end and back-end are part of the semiconductor manufacturing
fab and the use-phase is the final application in a railway power converter.

The front-end processing presents the highest energy consumption from all three stages [18, 19, 21, 22],
having the highest impact on the CED analysis of power devices. Due to its high complexity, one possible way of
obtaining such data is by industrial reports (when available) from the electricity use and the wafer total
consumption in a given period [19]. Our work used a direct measure of the fab energy consumption and the number
of fabricated chips. Hence, a metric of the energy required per wafer area can be accurately obtained for a specific
chip technology. Hitachi Energy Switzerland provided the manufacturing consumed energy data for Si and SiC
processed wafers. The Si IGBT data has a higher accuracy since this technology presents a well-established
production line.

The substrate production is the second stage in terms of cumulative energy demanded [18, 19, 21, 22].
The energy assessment in our work was performed for industrial standard monocrystalline 150 mm Si wafers and
150 mm 4H-SiC wafers. Intensive research and development on manufacturing larger wafers to reduce processing
costs and enhance chip production are being performed, with prospects for commercial 300 mm Si wafers [35]
and 200 mm SiC wafers [36]. The required data for the substrate CED were obtained from the literature [18, 19].
Regarding the embodied energy of the high-purity chemicals, gases and ultra-pure water used during the
processing (substrate, front-end and back-end), this category presents a lower impact on the CED analysis of
silicon semiconductor processing [19, 21, 22]. A contribution smaller than 11% of the total CED has been
consistently reported [18, 19, 21, 22]. Furthermore, the investigation of the used chemicals presents a high
uncertainty caused by the scarcity of semiconductor-grade chemicals' embodied energy data [18]. In order to keep
the modelling manageable and due to the relatively smaller contribution to the CED (as demonstrated for Si
processing), we have not considered the embodied energy of chemicals in the model [18].

6
Fig. 3 A. LinPak power module (Image from [34] - modified), B. Power module substrate (SiC version) with SiC
chips bonded on the copper (Image from [34] - modified).

Fig. 4 New generation of medium voltage power modules from different manufacturers. The modules are
compatible with Si and SiC chips.

The analysis provided in this paper mainly focuses on the substrate and front-end electricity requirements
due to their higher contribution to the total CED. Both stages also present the most significant differences between
different materials in the fabrication process. The technologies were compared during the manufacturing phase in
terms of the electrical energy consumed per wafer area (kWh/cm2). Finally, a comparison between Consumed
energy/module is provided, representing the module grey energy. We have considered the same assumptions from
[18] and [21] that the end-of-life energy requirement is negligible compared to substrate and front-end processing
and that the transport energy has a small contribution, being equivalent among distinct power semiconductors,
and not estimated in this work [18].

2.2. Use-phase in a railway application

In addition to the grey energy calculation, the energy losses during the use-phase of these power modules
in a railway traction application have been estimated. The comparison is performed in energy losses per power

7
module (kWh/module) in a train application running for 30 years, which is the typical required operating lifetime
for power converters in railway applications [37]. This analysis aims to assess the energy losses of the WBG
technology compared to the mainstream Si technology, considering the most energy-demanding steps of the power
module life-cycle.

The rail voltages can vary depending on the country [38, 39], ranging from 750 Vdc to 25 kVac.
However, even high AC voltages are reduced using onboard transformers and rectified to a DC link voltage of
750 V up to 3 kV for the traction inverter. Initially, we considered traction inverters used in metros, trains and
locomotives with a typical DC link voltage range of 1500 - 1800 V. The industry standard choice for such DC
link voltage is a three-phase two-level inverter topology, as demonstrated in Fig. 5, with 3.3 kV power switches
[26, 38]. The nominal power of these inverters is generally in the range of 150 kW – 1.4 MW [38], depending on
the power module rating and design considerations. Each leg of the 3-phase 2-level topology demonstrated in Fig.
5 is a LinPak power module composed of one power switch on the high side and one on the low side. Each switch
has a nominal current of 450 A, composed of several chips connected in parallel and bonded on a substrate (see
Fig. 3B). Considering the maximum allowed power module current, this power converter can theoretically reach
a nominal power rating of around 1 MW [27].

Nevertheless, power derating is required due to thermal limitations and to allow the long-term reliability
of the power modules [40]. The nominal power is generally based on 50 – 70 % of the power module nominal
current, depending on the design considerations (switching frequency, cooling design, expected module lifetime).
Thus, the traction converter with LinPak power modules may present a nominal power rating of around 500 - 700
kW. When higher traction power is required, the paralleling of power modules to increase converter power rating,
the use of higher current rating power modules or several power converters distributed through the carriages of
the train are possible adopted solutions [38].

The power loss evaluation of both technologies is based on a railway drive cycle profile from a real train
for suburban operations in central Europe. The semiconductor losses were simulated with software PLECs,
considering the datasheet semiconductor electro thermal parameters from 3.3 kV/450A Si IGBT and SiC
MOSFET LinPak power modules. Thus, the semiconductor losses during the acceleration, brake and cruise phases
could be assessed. Typical induction motor, train, and control parameters were considered in the simulation. The
lifetime energy losses were calculated with realistic considerations of the daily train operation time and the
simulation results.

8
Fig. 5 Three-phase two-level inverter topology connected to induction motors in the output. Each Half-bridge leg
corresponds to a LinPak power module composed of 2 power switches (high-side switch and low-side switch).

3. Power module manufacturing analysis

This section highlights the process steps for Silicon (Si) and Silicon Carbide (SiC) power modules and
their cumulative energy demand.

3.1. Silicon IGBT LinPak module

In power electronics applications, high-purity monocrystalline electronic-grade silicon is required. In

particular, contamination with elements that are also suitable, as dopants must be brought below critical values by
means of the Czochralski process [41]. This process consists in filling a rotating quartz crucible with the
polycrystalline silicon obtained in the Siemens process and melting it at a temperature higher than 1400 oC in an
inert gas atmosphere (e.g., argon). Due to its high heat conductivity, the quartz crucible sits inside a graphite
crucible, which homogeneously transfers the heat from the surrounding heater to the quartz crucible. A
monocrystalline silicon seed crystal with the desired crystal orientation is dipped into the melt. It acts as a starting
point for the crystal formation supported by the heat transfer from the melt to the already-grown crystal. A slight
temperature drop initiates the crystallization of silicon on the seed crystal. The seed crystal is slowly pulled upward
out of the melt, and a cylindrical silicon monocrystal hanging on the seed crystal starts to form. The pulling rate
and temperature are regulated such that the silicon monocrystal, whose orientation and structure are identical to
the seed crystal, can be pulled with a constant diameter. The diameter is usually around 150 to 300 mm, with a
height of around 2 m. High-purity silicon crystallizes in a monocrystalline form on the crystal, leaving almost all
impurities in the melt. The monocrystals grown with the Czochralski technique are cut into shorter workable
cylinders with a band saw. In order to increase the throughput, wire saws with many parallel wires, which cut
many wafers at once, are used. The wafers are then polished, as surface damage can occur on the wafer during
sawing.

9
 The analysis from electronic-grade Si substrates adopted in our work has already been estimated by [19]
with an adjusted yield of 75 % [18], providing an estimate of 0.45 kWh/cm2. The wafers are further processed at
into finished packs. They undergo more than 200 steps during the front-end stage. The main processes are listed
below:

Wafer Oxidation: This process is performed for gate oxide growth and to create a protective layer for other steps
during the front-end stage (e. g. ion implantation). The wafers are lined up and annealed in a furnace with the
addition of oxygen gas to grow an oxide film on the wafer surface.

Photolithography: This process aims to transfer the device patterns onto the wafer that will be further processed
in front-end steps. Around 12 different masks are used in an industrial IGBT manufacturing cycle. A
photosensitizing agent known as a photoresist is applied to the surface. The silicon wafer is rotated at high speed
during the coating process, so the photoresist is applied thinly and uniformly onto the wafer. A semiconductor
lithography system aligns the silicon wafer and the photomask, reduces the electronic device pattern on the
photomask and then projects light on the device pattern to the silicon wafer. After the light exposure, the
developing process is performed to dissolve the photoresist exposed areas (in the case of positive photoresists),
immersing the wafer in a liquid solvent to transfer the pattern onto the wafer’s surface.

Etching: The etching removes any unwanted material from the wafer during processing. This process is performed
several times and can be executed with liquid chemicals (wet etching) or in a plasma environment (dry etching),
depending on the desired etching pattern or target material.

Doping: Transistors and diodes must have different electrical conductivity values in particular device regions to
achieve the desired performance and functionality. External elements (dopants) such as boron, phosphorus, arsenic
and other elements are inserted in the crystalline structure of the wafer in a controlled way. Possible methods are
diffusion at high temperatures or ion implantation (ion bombarding).

Annealing: High-temperature processes in furnaces are required several times for material growth (e. g. wafer
oxidation), material deposition (e. g. polysilicon gate deposition), dopant electrical activation (after ion
implantation), dopant diffusion, and metal sintering to reduce electrical contact resistance.

Metallization: Metals are deposited onto the surface and in the back of the wafer to perform the electrical
connections of the fabricated device to the external circuit. Several metals can be deposited using physical
deposition methods (e. g. Sputtering).

The Si front-end stage was divided into two case scenarios, considering high and low utilization of the
production facility. Such cases occur because the energy demand from some manufacturing processes have a fixed
energy requirement. Consequently, the energy demand per wafer may vary according to the number of fabricated
devices. At a high utilization rate of the production plant (Scenario 1), the energy requirement is 183.9 kWh (per
150 mm (6") wafer). At a low utilization rate (Scenario 2), the energy consumed is higher, leading to an electrical
energy demand of about 221.8 kWh (per 150 mm (6") wafer). These data have been provided by the power
semiconductor fab of Hitachi Energy in Switzerland, for the IGBT module production line, encompassing the
global average energy demand of Si IGBTs and anti-parallel diodes. In calculating the energy demanded per area
for the front-end process, the total required wafer energy was divided by the number of chips per wafer and then

10
by the chip area. For a 6" wafer, sixty Si (60) chips of 1.3 x 1.3 cm each can be produced (typical IGBT and diode
3.3 kV chips dimension – information from an open module [42]), excluding losses, and fifty-seven (57) chips
when production losses are considered (yield of 95 %). The energy required for production per chip varies
depending on the capacity utilization of the production facility. Thus, the two scenarios were therefore calculated:
For 3.3 kV Si IGBT chips of dimensions 1.3 x 1.3 cm, with a yielding of 57 chips, we have a total front-end
energy consumption for the high utilization and low utilization plant use of 1.91 kWh/cm2 and 2.30 kWh/cm2,
respectively. A LinPak 3.3 kV/450 A half-bridge comprising two switches (high and low side switches
considered) requires a total of 24 Si chips. Considering the substrate and front-end energy demands, the Si LinPak
grey energy is 95.72 kWh and 111.54 kWh for the high (Scenario 1) and low (Scenario 2) utilization rate of the
production plant, respectively. The demanded energy per chip area and the module grey energy are indicated in
Tables 1 and 2, respectively. This work's calculated front-end energy demand (1.91 up to 2.3 kWh/cm2) is very
close to the estimated value from [18, 19] of 2 kWh/cm2. Such a fact indicates that using microelectronic front-
end estimates in power electronics Si IGBT devices does not significantly affect final accuracy.

Table 1 Substrate and Front-end energy demand estimate (kWh/cm2).

Manufacturing stage * Si devices SiC devices

Substrate 0.45 kWh/cm2 4.04 kWh/cm2
Front-end (Scenario 1) a 1.91 kWh/cm2 2.50 kWh/cm2
Front-end (Scenario 2) b 2.30 kWh/cm2 3.01 kWh/cm2
Front-end (Scenario 3) c 1.91 kWh/cm2 4.99 kWh/cm2
Front-end (Scenario 4) d 2.30 kWh/cm2 6.02 kWh/cm2
a
High production Si facility utilization with SiC 1.5 × Si energy demand.
b
Low production Si facility utilization with SiC 1.5 × Si energy demand.
c
High production Si facility utilization with SiC 3 × Si energy demand.
d
Low production Si facility utilization with SiC 3 × Si energy demand.
*
For Si devices, scenarios 1 and 3 are equal (High facility utilization) and scenarios 2 and 4 are equal (Low facility utilization).

Table 2 Estimate of the LinPak 3.3 kV/450 A module grey energy (kWh/module).

Cases * Si IGBT module SiC MOSFET module

a
Scenario 1 95.72 kWh 65.40 kWh
Scenario 2 b 111.54 kWh 70.50 kWh
c
Scenario 3 95.72 kWh 90.30 kWh
d
Scenario 4 111.54 kWh 100.60 kWh
a
High production Si facility utilization with SiC 1.5 × Si energy demand.
b
Low production Si facility utilization with SiC 1.5 × Si energy demand.
c
High production Si facility utilization with SiC 3 × Si energy demand.
d
Low production Si facility utilization with SiC 3 × Si energy demand.
*
For Si devices, scenarios 1 and 3 are equal (High facility utilization) and scenarios 2 and 4 are equal (Low facility utilization).

11
 The back-end demanded energy was not evaluated due to its minor contribution to the CED and for not
presenting significant energy demand differences between Si and SiC LinPak technologies, as explained in section
2.1. However, we will detail the main steps from this stage.

Dicing: The final wafer is further cut into individual chips. Generally, diamond blades are used to cut the chips
while spraying ultra-pure water.

Soldering: The individual chips are soldered to the substrates in a high-temperature chamber (~ 150 – 200 ⁰C). A
metallic alloy is melted to join the chip bottom part to the Cu substrate. This step aims to fix the chips to the
substrate and to electrically connect the bottom chip contact (MOSFET Drain and IGBT Collector) to the
substrate.

Wire bonding: The top chip electrical connections are bonded with wires to the copper substrate with an ultrasonic
wire bonder machine. This step electrically connects the top chip contacts (Gate, MOSFET Source and IGBT
Emitter) to the substrate.

Packaging: The substrates are soldered to the base plate and fixed inside a closed plastic case. Internal metallic
connections are used to connect the substrate to the external connectors. The module is filled with silicone gel for
improved electrical isolation, mechanical stability, reliability and dust protection.

3.2. Silicon Carbide MOSFET LinPak module

SiC processing is expected to present higher energy demand than silicon in substrate and front-end
processing. Such a fact happens mainly due to the higher processing complexity that generally requires higher
temperatures. SiC substrate fabrication is mainly performed with the modified Lely method, being the industry
standard due to the good wafer quality and high yield (90%) [43]. This process is based on a sublimation growth
technique, where the source material is held at the bottom of the crucible and the seed plate on the top. The
environment temperature is in the range of 1800 – 2600 oC with argon gas at 10-4 to 760 torr, which the seed is
kept at a slightly lower temperature than the source to facilitate the vapor transport [43]. In our work, we have
used the estimate from [18] for the modified Lely process given as 4.04 kWh/cm2.

The front-end stage has almost the same steps as the Si processing, with different recipes for each stage.
SiC requires a temperature higher than 1600 oC [6, 44] for dopant activation, whereas Si uses a temperature
between 450 – 750 oC [45]. Furthermore, the gate oxide growth in SiC also requires a slightly higher temperature
than Si, with a temperature range between 1100 – 1400 oC [44] and Si in the range of 900 – 1200 oC [46]. Such
higher temperatures make SiC require a higher energy demand than Si during the front-end. The other processes
require lower energy since high temperature is not used, presenting similar magnitudes between both materials,
as shown in Fig. 6. For a 6” wafer, excluding production losses, 520 chips with dimensions of 0.5 x 0.5 cm each
(typical MOSFET 3.3 kV chips dimension [47]) can be produced. Including process losses, the value reduces to
442 (yield of 85%). The yield value for the SiC technology is very dependent on the wafer quality as well as the
power semiconductor manufacturer design and front-end processes, presenting a significant variation between
different fabs. Such value also increases with further technology development in substrate fabrication and front-
end processing techniques. The value of 69 % was assumed by Ref [18] 8 years ago, and further developments
have been performed during these years. We estimate an 85 % yield is a realistic value that can be achieved in the

12
future when these modules become a standard commercial product due to intensive improvements in SiC
processing.

The SiC front-end energy data from Hitachi Energy have been provided in terms of range values
comparative to Si. The front-end production of SiC MOSFETs requires 1.5 to 3 times more energy per 6” wafer
than Si IGBTs. The large range is due to the different possibilities of processes involved in the 3.3 kV SiC device
production, and accounts further for uncertainties reflected on a SiC production line incorporated within a Si
manufacturing facility. It is worth mentioning, that implementing SiC into a high volume Si facility can strongly
reduce manufacturing costs. In addition, these values are also dependent on the automation level of the
manufacturing site, but since annealing processes are expected to be the main driven for the SiC higher
consumption, the trend is expected to remain similar. The SiC front-end analysis was then split into 4 case
scenarios (shown below), considering high and low utilization of the production facility and the range limits (1.5
and 3 X).

Scenario 1: × 1.5 the energy demand of silicon production with high production Si facility utilization.

Scenario 2: × 1.5 the energy demand of silicon production with low production Si facility utilization.

Scenario 3: × 3 the energy demand of silicon production with high production Si facility utilization.

Scenario 4: × 3 the energy demand of silicon production with low production Si facility utilization.

The calculated demanded energy per area (kWh/cm2) for the front-end stage and module grey energy
(kWh) for all scenarios are demonstrated in Tables 1 and 2, respectively. Fig. 7 presents a graphical representation
of the power modules' grey energy with the substrate and front-end contributions indicated for each scenario. This
graphical comparison shows that for SiC technology, the substrate production may present a higher CED than the
front-end stage, depending on the plant utilization.

It has been considered that one wafer of 6” can generate 442 chips of 0.5 x 0.5 cm each requiring a total
of 40 chips in the 3.3 kV/450 A LinPak power module [27]. The SiC MOSFET module presents a lower CED
than the Si IGBT module in all possible scenarios. This fact occurs even considering that the SiC chip yield is
lower than the Si IGBT (85 % versus 95 %), caused by the larger number of defects in the SiC wafer. On top of
that, the SiC substrate energy demand per area is also higher than the Si substrate (4.04 versus 0.45 kWh/cm2) as
well as the front-end stage. In fact, the dominant factor is the smaller area per Ampere required per chip in the
SiC technology compared to Silicon. Additionally, using the internal body diode as an antiparallel diode [47], not
requiring additional external diode chips as in the case of Si IGBTs, also reduces the required area. As a result, a
much smaller area for the same current rating (450 A) is required (10 versus 40.56 cm2). This fact occurs due to
the SiC intrinsic characteristics [1], allowing the design of high-power density chips. The SiC LinPak grey energy
may present further reduction in the future with the implementation of a High-temperature CVD process [18].
Such a method can potentially reduce the wafer defect density [18], attracting semiconductor manufacturers for
further research. An estimate of 2.78 kWh/ cm2 demanded energy for this process is given in Ref [18], which can
reduce the SiC LinPak substrate energy demand to a value of 27.8 kWh compared to the Lely method of 40.4
kWh.

13
Fig. 6 Energy demand representation from different front-end processes for Si and SiC. Dopant activation and
oxide growth are the processes with the highest thermal budget required, with SiC presenting the highest values.

Regarding the front-end estimates, our values are in the range of 2.50 up to 6.02 kWh/cm2, depending on
the chosen scenario. Ref [18] estimated 2.43 kWh/cm2 based on 600 V SiC Schottky diode processing data. Our
estimates are about 1.03 – 2.48× higher than the estimate of [18]. Such difference may be related to the fact that
3.3 kV SiC MOSFETs may present a higher energy demand than 600 V SiC Schottky diodes since low voltage
diodes present more straightforward manufacturing steps.

Fig. 7 Power modules grey energy for different scenarios with the substrate and front-end energy demand
contributions indicated separately.

14
4. Use-phase analysis

The energy assessment of the SiC technology was performed under a realistic drive cycle from a real
train for suburban operations in central Europe, as shown in Fig. 8A. The route has a distance of 11 km completed
in 384 s (6.4 min). The maximum speed during cruising is 38.89 m/s (140 km/h), and the maximum accelerations
are 0.66 m/s2 and 1 m/s2 at the acceleration and brake phases (see Fig. 8B), respectively. It has been assumed that
this drive cycle was performed in a flat route without curves. We have focused on simulating one commercial
power converter composed of three 3.3kV/450 A LinPak power modules (Fig. 5) connected to 4 medium voltage
induction motors, which is a typical situation in actual trains where from 2 up to 4 motors can be connected to the
same inverter output [38]. The power converter has a nominal power rating of around 500-700 kW, considering
the power derating as explained in section 2.2. The electrical and thermal behavior of the Si and SiC LinPak
modules are modelled based on datasheet and literature data [27, 32]. The modelling of the conduction and
switching losses consider the instantaneous device junction temperature, current and voltage. These considerations
were implemented by interpolating forward and switching loss characteristics at different temperatures (25 and
125 ⁰C) during each simulation step. A water cooler was considered with a thermal resistance from the base plate
to the water of 40 K/kW and a constant water temperature of 50 ⁰C. A typical switching frequency of 600 Hz from
3.3 kV Si IGBT modules was used. The four motors have a nominal power of 640 kW (each 160 kW) and present
a nominal voltage of 1287 V. This value is within the range of a typically adopted motor voltage for converters
with a DC link of 1500-1800 V [38]. The electrical parameters of the motor used in the simulation are detailed in
[48]. The chosen control strategy is the rotor flux-oriented vector control, one traditional technique for the high-
performance control of induction motor drives in railway applications [38, 49]. Based on the mechanical vehicle
parameters, the motor torque and motor speed are obtained for the chosen mission profile. From the motor
electrical parameters, the direct and quadrature voltage and current components are calculated based on the
dynamic equations of the IM machine [38]. The voltage values can be used to generate the reference signal to the
PWM modulator of the PLECS electrothermal model to simulate the drive cycle profile with thermal modelling
included. The output current and motor speed can be measured and compared to the previously calculated values
to check the accuracy of the simulation with the theoretically calculated values, presenting an excellent match.
Further details on the equations used are shown in Ref [38].

Fig. 8 A. Drive cycle from a suburban train route in central Europe, B. Train acceleration during the drive cycle.

15
 The train parameters and modelling are based on the work of [50], considering the most significant force
contributions. Equations 1-3 show the parameters for the requested motor force.

𝐹 = 𝑚. 𝑎 (1)

𝐹𝑑 = (0.75 + 0.0011664 ∗ 𝑣 2 ) ∗ 𝐺 (2)

𝐹𝑡 = 𝐹 + 𝐹𝑑 (3)

The requested motor force (Ft[N]) is the sum of the inertial (F[N]), and drag (Fd[N]) forces. The other
parameters from Eqs. 1-3 are train acceleration (a[m/s2]), train weight (G [kN]) and train speed (v[m/s]). The
considered coefficients are from modern passenger trains [50]. Some additional suitable parameters are
considered, such as the wheel diameter of 0.687 m [51] and a gear ratio of 1 to cover the vehicle speed range from
Fig. 8A, according to the nominal motor parameters. Since we focus on the power module losses, just one power
converter must be simulated, reducing computer simulation effort. In a real train, several power converters and
motors are connected to the train's carriages (see section 2.2), and the traction effort is equally divided under ideal
conditions. Thus, the effective train mass (m) of 12 Tons was considered to simplify the simulation and obtain the
practical effect on one power converter. This value corresponds to the effective effort distributed to the simulated
power converter and not from the whole train. This value was chosen so that the maximum power converter
current during operation reaches the derated nominal current (~ 50 – 70% of the datasheet nominal current – Point
C from Fig. 8), which is the typical current allowed due to reliability, cooling and module lifetime considerations
(see section 2.2). Fig. 9 shows the inverter output current in one output phase for three distinct moments indicated
in Fig. 8 (Points A, B, and C). The current waveforms are typical of a sinusoidal pulse width modulation (SPWM)
modulation technique with variable frequency in function of the shaft speed. As expected, the ratio between the
switching frequency and output frequency decreases for higher shaft speeds and the current distortion increases
[51]. As one can see, point B presents the highest distortion, followed by C and A. Another characteristic is the
high current value during acceleration and brake phases caused by the inertial force component that requires most
of the motor traction effort. Point C, with the highest acceleration value of – 1m/s2, presents the highest
fundamental rms current of 290 A, followed by points A (~ 205 A) and B (~150 A).

Fig. 9 A. One-phase inverter output current at point A from Fig. 8, B. One-phase inverter output current at point
B from Fig. 8, C. One-phase inverter output current at point C from Fig. 8.

The details of the Si and SiC based power converters simulated in software PLECs are shown in Table
3. One can see that the power converters have the same nominal power, with the SiC converter presenting a 4 ×

16
lower chip area and reduced grey energy, considering only the power modules’ grey energy. Alternative designs
can be made, such as an increase in the switching frequency to improve power density and reduce the current
harmonic distortion. Such choice come at the expense of reduced power conversion efficiency. We have
considered the same switching frequency and power rating limited by the module nominal current [25].

Table 3 3-phase inverter details and comparison between manufacturing and use-phase energy requirements for
Si and SiC based converter.

Si SiC
Converter rating [kW] 500 – 700 500 – 700
Converter chip area [cm2] 121.7 30
Converter grey energy [kWh]* 335 302
Switching frequency [Hz] 600 600
Lifetime CED [MWh] 181.9 109.9
Net life cycle energy change from Si [MWh] -72
*
Considering only the power module’s grey energy. Passive components, busbar, gate drives, cooling, metal case are not considered. Scenario

4 was considered for the power modules grey energy.

The simulation results show that one SiC module (phase-leg) could save around 40 % of the electric
energy loss of a Si IGBT module, with the Si IGBT and SiC MOSFET modules presenting energy losses of 0.0623
kWh and 0.0376 kWh per module (per phase) in one drive cycle, respectively.

We assume a typical train operation of about 12 h per day, during which the drive cycle will be repeated
around 108 times daily (drive cycle duration approximated to 400 s). Considering a train operation of 300 days
per year, the Si IGBT and SiC MOSFET power module will present a yearly energy loss of 2.018 MWh and 1.218
MWh, respectively. In a 30 years lifetime, the Si IGBT and SiC MOSFET module energy losses are 60.54 MWh
and 36.54 MWh, respectively. Considering scenario 4 for the module’s grey energy, the lifetime CED are 60.65
MWh and 36.64 MWh for the Si IGBT and SiC MOSFET modules, respectively. Figure 10 shows the energy
demand per power module during the entire life cycle, considering module grey energy and use phase. The saved
energy per module during the power module lifetime with the substitution of Si to SiC technology is around 24
MWh. Independently of the chosen scenario for the power module grey energy, the use-phase is the dominant
stage in terms of consumed energy, with the manufacturing stage presenting a negligible contribution to the entire
power module life cycle (smaller than 0.3 %). This fact agrees with previous WBG investigations in electric car
applications [18] and LED devices [52, 53], where the use-phase energy consumption dominates by several orders
of magnitude. A consequence of the minor grey energy contribution to the entire lifecycle is that in case module
replacements are necessary during the converter lifetime, such replacements present negligible influence on the
potential energy saving of SiC technology. In our simulation, we considered a typical switching frequency of 600
Hz from 3.3 kV Si IGBT to perform a comparison at the same operating conditions [25]. However, depending on
the design specifications desired by the converter manufacturer, higher switching frequencies up to around 1.5
kHz [8] can improve the current harmonic distortion and consequently improve motor efficiency. The optimum
switching frequency depends on the motor design and manufacturer design considerations, presenting significant
variations between manufacturers. In order to consider the limit cases from an inverter point of view, a simulation

17
at 1.5 kHz for the SiC module was performed, providing energy losses of 0.0391 kWh per module (per phase) in
one drive cycle. With the same assumptions of a typical train operation performed before, the saved energy per
module during the power module lifetime with the substitution of Si to SiC technology reduces to a value of 22.6
MWh. Such reduction in the energy saving potential is smaller than 6 % from an operation at 600 Hz (24 MWh).
It is essential to reinforce that this analysis is limited to the power module losses. Further research is necessary to
investigate the energy saved from the motor side with the current harmonic distortion reduction at a higher
switching frequency. Such analysis requires a global system investigation, which is not the focus of this paper.

Fig. 10 Lifetime energy demand of Si and SiC LinPak power modules. The grey energy contribution to the
lifetime CED is smaller than 0.3 %.

The results demonstrated here partially contradicts the conclusions of [18] that power semiconductor
manufacturers may have to increase manufacturing energy use to save many times more energy in the use-phase
stage. Actually, the energy demanded per wafer area is higher for WBG technology. However, the SiC power
module, the basic unit cell of a power converter, presents a lower grey energy, indicating that the energy demand
is smaller than Si when comparing equal current rating devices. In the considerations of [18], inverters with the
same semiconductor area for Si and SiC technologies have been considered, which may lead to an unfair
comparison of converters with very different power ratings. Another advantage of the SiC technology, not
investigated in this work, is the power converter size and weight reduction [7, 31, 54] that may also contribute to
reducing the manufacturing energy spent on power converters due to passive components, cooling and metal case
size reduction.

18
Fig. 11 Payback time curves per module during the inverter’s lifetime. A. Railway electricity cost of 47.35
cents per kWh, B. Railway electricity cost of 13 cents per kWh.

A last investigation is the payback time of substituting Si IGBT with SiC MOSFET modules. The Silicon
IGBT LinPak module costs around 1000 € based on market quotations. Since the SiC LinPak module is not
available on the market yet, an assumption based on lower voltage commercial devices is that this module will
cost around three times more than the silicon version, leading to a value of 3000 €. The railway electricity cost
can present a huge variation between countries and tends to increase even more in the following years [55, 56].
Germany recently imposed a price of 13 cents per kWh for 2023 to alleviate the skyrocketing prices affecting
railway transportation [55]. France presents for 2023 a cost of 47.35 cents per kWh [56]. Both cases have been
considered to evaluate the payback time of SiC in railway inverters. Fig. 11A shows that for the case of 47.35
cents per kWh, after five years, the cumulative cost of SiC and Si are equalized, and at the end of the inverter
lifetime, a cost saving per module of 9364 €. For the case of lower energy cost, Fig. 11B shows that the time
required to equalize the cumulative cost is around 19 years, with a cost saving per module at the end of the inverter
lifetime of 1120 €. Our analysis is performed for both power module technologies operating at the same switching
frequency (600 Hz) and considering no module replacement during the whole lifetime. In the case of higher
switching frequency designs, the SiC module suffers a slight increase in losses. However, further advantages on
the motor side can be achieved regarding energy savings. Another consideration is whether module replacements
are necessary for maintenance during the converter's lifetime. In such a case, a maximum of 3 replacements
(considering the same number of replacements for both technologies) can be performed during the inverter lifetime
to keep SiC financially advantageous for the case of high energy cost (47.35 cents/kWh). In the case of a lower-
cost energy scenario (13 cents/kWh), any replacement places silicon technology with a financial advantage. It is
important to reinforce that the values used are based on cost assumptions. Manufacturers can buy a large number
of semiconductor devices with discounts, presenting further reductions in the cost difference between both
technologies, favouring the SiC technology. With the strong tendency of railway electricity costs to increase in
the following years and further advances in SiC technology, SiC MOSFETs become even more appealing in terms
of improved energy and cost saving potential.

19
5. Conclusions

This work estimated the potential energy savings of implementing SiC technology compared to Si
technology in railway traction converters, considering the power module life cycle in an operational railway
inverter lifetime of 30 years. The manufacturing energy per area was calculated based on industrial data and
literature analysis. Even though SiC presents a 2.6 – 3.8× higher manufacturing CED per wafer area, the power
module grey energy was smaller than Si IGBT modules (around 1.1 – 1.6× smaller). This fact happens due to the
enhanced power density and the non-use of additional antiparallel diodes in SiC MOSFET power modules. Both
characteristics make SiC power modules require a lower wafer area than Si for the same current (power rating)
capacity. This paper also aims to fill a gap in the available literature data, providing front-end data from a power
semiconductor fab that manufactures both technologies. Researchers could then use such data as a baseline to
perform life cycle energy investigations of several power electronic systems, not limited to 3.3 kV systems. Some
examples are transport electrification and renewables, which may present a substantial global energy impact with
Si and SiC technologies. Furthermore, this data may be a reference for the power semiconductors industry to
investigate possible manufacturing process optimizations.

The use-phase analysis in a typical railway drive cycle shows that SiC can save up to around 40 % of the
energy losses from Si IGBT modules. Furthermore, the manufacturing energy contribution to the lifetime CED is
negligible (contribution smaller than 0.3 %), with the use-phase energy demand being the most critical. Finally,
considering the whole power module life-cycle, around 24 MWh/module can be saved with SiC implementation
in railway traction converters. We believe the analysis provided in this paper can help decision-makers about the
environmental implications of such technologies to help formulate governmental policies. With further
development of the SiC technology, the costs may decrease, and increased reliability may be achieved, favouring
the implementation of such technology to allow more efficient future power systems. Some actual train lines
already present 3.3 kV SiC power modules in their converters, and more trains are expected to implement such
technology in the future for more efficient railway transport.

Acknowledgements The authors thank Hitachi Energy, Switzerland for the manufacturing energy data. This
work was supported by the Swiss Federal Office of Energy SFOE under SiC MILE project (Grant No. SI/502290).

Conflict of Interest: The authors declare that they have no conflict of interest.

References

1. Yuan X, Laird I, Walder S. Opportunities, Challenges, and Potential Solutions in the Application of Fast-
Switching SiC Power Devices and Converters. IEEE Transactions on Power Electronics, 2021, 36(4):3925-3945

2. Blaabjerg F, Round S - Power Electronics: Revolutionizing the World’s Future Energy Systems. Hitachi
Energy. 2021, available at https://www.hitachienergy.com/news/perspectives/2021/08/power-electronics-
revolutionizing-the-world-s-future-energy-systems. Accessed 03 February 2023

3. Simiyu P, Davidson I E. MVDC Railway Traction Power Systems; State-of-the-Art, Opportunities, and
Challenges. Energies, 2021, 14(4156):1-27

20
4. Li L, Zhang X. Integrated optimization of railway freight operation planning and pricing based on carbon
emissions reduction policies. Journal of Cleaner Production, 2020, 263(121316):1-13

5. Ding X, Du M, Zhou T, Guo H, Zhang C. Comprehensive comparison between silicon carbide MOSFETs and
silicon IGBTs based traction systems for electric vehicles. Applied Energy, 2017, 194:626-634

6. Li F, Jennings M. Main Differences in Processing Si and SiC Devices. In: Sharma Y K, ed. Disruptive Wide
Bandgap Semiconductors, Related Technologies, and Their Applications. 1st ed. Chapter 3, 2018, available at
https://www.intechopen.com/chapters/60647

7. Sato K, Kato H, Fukushima T. Development of SiC Applied Traction System for Next-Generation Shinkansen
High-Speed Trains. IEEJ Journal of Industry Applications, 2020, 9(4):453-459

8. Rujas A, Lopez-Martin V M, Jauregi A, Larzabal I, Nieva T. Railway inverter for metro application with 3.3kV
Full-SiC MOSFET modules. In: 2021 IEEE Vehicle Power and Propulsion Conference (VPPC), 2021, Doi:
10.1109/VPPC53923.2021.9699286

9. Siemens. Siemens Mobility and Mitsubishi Electric Europe B. V. sign MoU to push the use of high voltage SiC
technology for increasing the energy efficiency of railway drive systems. Siemens, 2022, Available at
https://press.siemens.com/global/en/pressrelease/siemens-mobility-and-mitsubishi-electric-europe-bv-sign-mou-
push-use-high-voltage-sic. Accessed 11 April 2023

10. Nordelöf A, Alatalo M, Söderman M L. A scalable life cycle inventory of an automotive power electronic
inverter unit-part I: design and composition. The International Journal of Life Cycle Assessment, 2019, 24:78-92

11. Nordelöf A. A scalable life cycle inventory of an automotive power electronic inverter unit-part II:
manufacturing processes. The International Journal of Life Cycle Assessment, 2019, 24:694-711

12. Koroma M S, Costa D, Phillipot M, Cardellini G, Hosen M S, Coosemans T, Messagie M. Life cycle
assessment of battery electric vehicles: Implications of future electricity mix and different battery end-of-life
management. Science of the Total Environment, 2022, 831(154859):1-12

13. Vilaça M, Santos G, Oliveira M S A, Coelho M C, Correia G H A. Life cycle assessment of shared and private
use of automated and electric vehicles on interurban mobility. Applied Energy, 2022, 310(118589):1-12

14. Jungbluth N, Bauer C, Dones R, Frischknecht R. Life Cycle Assessment for Emerging Technologies: Case
Studies for Photovoltaic and Wind Power. The International Journal of Life Cycle Assessment, 2005, 10(1):24-
34

15. Ali A, Koch T W, Volk T A, Malmsheimer R W, Eisenbies M H, Kloster D, Brown T R, Naim N, Therasme
O. The Environmental Life Cycle Assessment of Electricity Production in New York State from Distributed Solar
Photovoltaic Systems. Energies, 2022, 15(7278):1-20

16. Yu Z, Ma W, Xie K, Lv G, Chen Z, Wu J, Yu J. Life cycle assessment of grid-connected power generation

from metallurgical route multi-crystalline silicon photovoltaic system in China. Applied Energy, 2017, 185(1):68-
81

21
17. Garcia-Teruel A, Rinaldi G, Thies P R, Johanning L, Jeffrey H. Life cycle assessment of floating offshore
wind farms: An evaluation of operation and maintenance. Applied Energy, 2022, 307(118067):1-17

18. Warren J A, Riddle M E, Graziano D J, Das S, Upadhyayula V K K, Masanet E, Cresko J. Energy Impacts of
Wide Band Gap Semiconductors in U.S. Light-Duty Electric Vehicle Fleet. Environmental Science &
Technology, 2015, 49(17):10294-10302

19. Williams E D, Ayers R U, Heller M. The 1.7 kilogram microchip: Energy and material use in the production
of semiconductor devices. Environmental Science & Technology, 2002, 36(24):5504-5510

20. Schmidt M, Hottenroth H, Schottler M, Fetzer G, Schlüter B. Life cycle assessment of silicon wafer processing
for microelectronic chips and solar cells. The International Journal of Life Cycle Assessment, 2012, 17:126-144

21. Boyd S B, Horvath A, Dornfeld D. Life-cycle energy demand and global warming potential of computational
logic. Environmental Science & Technology, 2008, 43(19):7303-7309

22. Krishnan N, Boyd S, Somani A, Raoux S, Clark D, Dornfeld D. A hybrid life cycle inventory of nano-scale
semiconductor manufacturing. Environmental Science & Technology, 2008, 42(8):3069-3075

23. Murphy C F, Kenig G A, Allen D T, Laurent J, Dyer D E. Development of parametric material, energy, and
emission inventories for wafer fabrication in the semiconductor industry. Environmental Science & Technology,
2003, 37(23):5373−5382

24. Díaz Triana A, Schmidt S, Glaser S, Makoschitz M. A “life cycle thinking” approach to assess differences in
the energy use of SiC vs. Si power semiconductors. In: e.nova 2021 Conference, 2021

25. Soltau N, Wiesner E, Stumpf E, Idaka S, Hatori K. Electric-Energy Savings using 3.3 kV Full-SiC Power-
Modules in Traction Application. In: 2020 15th International Conference on Ecological Vehicles and Renewable
Energies (EVER), 2020. Doi: 10.1109/EVER48776.2020.9242996.

26. Hruska M, Bhatnagar P, Sleven M. Benefits of Using the 1700V and 3300V High Power Modules for Traction
Applications. In: PCIM Europe 2021, 2021, pp 253-258.

27. Kicin S, Burkart R, Loisy J-Y, Canales F, Nawaz M, Stampf G, Morin P, Keller T. Ultra-Fast Switching 3.3kV
SiC High-Power Module. In: PCIM Europe 2020, 2020, pp 150-157

28. Marzoughi A, Burgos R, Boroyevich D. Investigating Impact of Medium-Voltage SiC Mosfets on Medium-
Voltage High-Power Industrial Motor Drives. IEEE Journal of Emerging and Selected Topics in Power
Electronics, 2019, 7(2):1371-1387

29. Bergmann L, Wahle M, Bakran M. Optimization of the chip area in 3.3 kV SiC submodules for HVDC
converters. In: 2021 23rd European Conference on Power Electronics and Applications (EPE’21 ECCE Europe),
2021, Doi: 10.23919/EPE21ECCEEurope50061.2021.9570447

30. Morel F, Stackler C, Ladoux P, Fouineau A, Wallart F, Evans N, Dworakowski P. Power electronic traction
transformers in 25 kV / 50 Hz systems: Optimisation of DC/DC Isolated Converters with 3.3 kV SiC MOSFETs.
In: PCIM Europe 2019, 2019, pp 553-560

22
31. Ding R, Dou Z, Qi Y, Mei W, Liu G. Analysis on characteristics of 3.3-kV full SiC device and railway traction
converter design. IET Power Electronics, 2022, 15:978-988

32. Hitachi Energy. Datasheet IGBT LinPak Module (3.3 kV/450 A) 5SNG 0450X330300. Available at
https://search.abb.com/library/Download.aspx?DocumentID=5SYA%201456&LanguageCode=en&DocumentP
artId=&Action=Launch. Accessed 07 March 2023

33. Zhang Y, Wu R, Iannuzzo F, Wang H. Aging investigation of the latest standard dual power modules using
improved interconnect technologies by power cycling test. Microelectronics Reliability, 2022, 138(114740):1-8

34. Hitachi Energy. Product Note: SiC LinPak. Hitachi Energy Switzerland Ltd. Semiconductors. Available at
https://search.abb.com/library/Download.aspx?DocumentID=9AKK107991A0985&LanguageCode=en&Docu
mentPartId=&Action=Launch. Accessed 07 February 2023

35. Deboy G, Aigner K. Power Semiconductors on 300-Milimeters Wafers. Power Electronics Europe, 2013,
Issue 3(April/May 2013): 44-47

36. Musolino M, Xu X, Wang H, Rengarajan V, Zwieback I, Ruland G, Crippa D, Mauceri M, Calabretta M,

Messina A. Paving the way toward the world’s first 200mm SiC pilot line. Materials Science in Semiconductor
Processing, 2021, 135(106088):1-5

37. Piton M, Batista E, Escrouzailles V. Improved Stress Distribution in Railway Traction Converters Using New
High Power Half-Bridge Modules. In: CIPS 2020, 11th International Conference on Integrated Power Electronics
Systems, 2020

38. Agirre X, Abad G. Railway traction. In: Abad G (ed) Power Electronics and Electric Drives for Traction
Applications, 2017, 1st edn. Chapter 5, Wiley, UK, pp 221-361

39. Drofenik U, Canales F. European Trends and Technologies in Traction. In: 2014 International Power
Electronics Conference (IPEC-Hiroshima 2014 – ECCE ASIA), 2014, Doi: 10.1109/IPEC.2014.6869715

40. Vernica I, Ma K, Blaabjerg F. Optimal Derating Strategy of Power Electronics Converter for Maximum Wind
Energy Production with Lifetime Information of Power Devices. IEEE Journal of Emerging and Selected Topics
in Power Electronics, 2018, 6(1):267-276

41. Eranna G. In: Crystal Growth and Evaluation of Silicon for VLSI and ULSI, 2015, CRC Press, Florida

42. Hitachi Energy. Datasheet IGBT HiPak Module (3.3 kV/1200 A) 5SNA 1200G330100. Available at
https://search.abb.com/library/Download.aspx?DocumentID=5SYA156303&LanguageCode=en&DocumentPart
Id=&Action=Launch. Accessed 27 February 2023

43. Dhanaraj G, Huang X R, Dudley M, Prasad V, Ma R-H. Silicon Carbide Crystals: Part I: Growth and
Characterization. In: Scheel HJ, Fukuda T (Eds.) Crystal Growth Technology, 2003, Chapter 6, William Andrew,
New York, pp 181-232

44. Langpoklakpam C, Liu A, Chu K, Hsu L, Lee W, Chen S, Sun C, Shih M, Lee K, Kuo H. Review of Silicon
Carbide Processing for Power MOSFET. Crystals, 2022, 12(245):1-27

23
45. Borland J O. Low temperature activation of ion implanted dopants: a review. In: Extended Abstracts of the
Third International Workshop on Junction Technology 2002, Doi: 10.1109/IWJT.2002.1225211

46. Jaeger R C. Introduction to Microelectronic Fabrication, 2002, Volume V, Second edition, Prentice Hall, New
Jersey

47. Knoll L, Mihaila A, Bauer F, Sundaramoorthy V, Bianda E, Minamisawa R, Kranz L, Bellini M, Vemulapati
U, Bartolf H, Kicin S, Skibin S, Papadopoulos C, Rahimo M. Robust 3.3kV Silicon Carbide MOSFETs with
Surge and Short Circuit Capability. In: Proceedings of the IEEE 29th International Symposium on Power
Semiconductor Devices and ICs (ISPSD), 2017, pp. 243-246, Doi: 10.23919/ISPSD.2017.7988905

48. Tang J, Yang Y, Blaabjerg F, Chen J, Diao L, Liu Z. Parameter Identification of Inverter-Fed Induction
Motors: A Review. Energies, 2018, 11(2194):1-21

49. Abouzeid A F, Guerrero J M, Endemaño A, Muniategui I, Ortega D, Larrazabal I, Briz F. Control Strategies
for Induction Motors in Railway Traction Applications. Energies. 2020, 13(700):1-22

50. Brenna M, Foiadelli F, Zaninelli D. Electrical Railway Transportation Systems. IEEE Press, 2018, John Wiley
& Sons, New Jersey

51. Spina I, Del Pizzo A, Beneduce L, Cascone B, Fratelli L. Comparative analysis of performance and energy
losses in light railways vehicles equipped with IM or PMSM drive. In: 2014 International Symp. on Power
Electron., Electrical drives, Automation and Motion, 2014. Doi: 10.1109/SPEEDAM.2014.6871998

52. Scholand M J, Dillon H E. Life-Cycle Assessment of Energy and Environmental Impacts of LED Lighting
Products-Part 2: LED Manufacturing and Performance. Solid-State Lighting Program, Building Technologies
Program, Office of Energy Efficiency and Renewable Energy, U.S. Department of Energy, 2012.

53. Osram Opto Semiconductors GmbH. Executive summary: Life cycle assessment of illuminants-A Comparison
of Light Bulbs, Compact Fluorescent Lamps and LED Lamps, 2009, Available at
http://seeds4green.net/sites/default/files/OSRAM_LED_LCA_Summary_November_2009.pdf. Accessed 10
March 2023.

54. Hamada K, Hino S, Miura N, Watanabe H, Nakata S, Suekawa E, Ebiike Y, Imaizumi M, Umezaki I,
Yamakawa S. 3.3 kV/1500 A power modules for the world’s first all-SiC traction inverter. Japanese Journal of
Applied Physics, 2015, 54(4S):04DP07

55. Official: Germany pumps 1.1 billion to cover rail electricity bills. Available at
https://eress.eu/news/news/official-germany-pumps-1-1-billion-to-cover-rail-electricity-bills/ Accessed 28
September 2023

56. French electricity price for rail will increase 745 %. Available at
https://www.railtech.com/all/2022/12/19/french-electricity-price-for-rail-will-grow-by-850-percent/ Accessed
28 September 2023

24
Graphical abstract

Declaration of interests

☒ The authors declare that they have no known competing financial interests or personal

relationships that could have appeared to influence the work reported in this paper.

☐The authors declare the following financial interests/personal relationships which may be
considered as potential competing interests:

25