**1. Introduction**

Lithium-ion batteries have become the foundation of a wide variety of applications depending on electrochemical energy storage [1], starting with small single cells in billions of mobile phones [2] to a vigorously growing amount of electric vehicles [3] and large battery storage power stations [4], although the latter is oftentimes viewed critically under economic aspects [5]. Regardless of the size or scope, all lithium-ion batteries have in common the need to be meticulously monitored in order to ensure safe and durable operation. This type of electrochemical energy storage only has a small range of tolerated operational states [6] and tend toward exothermic reactions, usually referred to as thermal runaway (TR) [7], if the limits for safe operation are violated.

If voltage or temperature specifications are violated, the resulting fire can quickly propagate through densely packed battery systems, causing the TR of neighbored cells and ultimately of the whole system. A number of accidents have already attracted public attention in the past [8] and it is safe to say, that, with increasing number of electric cars and high-powered stationary applications this attention will increase.

Additional to ensuring operational safety, adequate status monitoring enables the user to decide whether a lithium-ion battery may be used any further [9]; for example, a used traction battery with an

insufficient capacity may still operate fine in low power stationary applications. Extended monitoring of lithium-ion batteries therefore can affect the economic treatment of used batteries [10] as e.g., either waste or valued energy storage systems for stationary applications.

State-of-the-art monitoring systems for lithium-ion batteries consist almost exclusively of electronic battery managemen<sup>t</sup> systems (BMS) that obtain their information from electrical sensors [11]. The most frequently obtained measurement quantities are voltage, temperature and current. While voltage is usually acquired for each cell, current and temperature are often measured for modules and/or whole battery packs. A wide variety of methods for determining the state of charge (SOC) is currently utilized ranging from simple charge counting over model-based observers to neural networks and fuzzy logic algorithms. All methods yield a SOC accuracy equal or below 6% although diminishing capacity due to wear on the battery is not modeled in SOC estimation. State of health (SOH) estimation is therefore carried out in parallel fashion [12,13], using e.g., different battery models and the results are fed into the mostly adaptive SOC algorithms.

However, most applied methods still lack on accuracy and a safety margin for allowed battery SOC and SOH is usually applied to account for uncertainties and avoid financial damages e.g., due to warranty claims.

Additional information about the SOH of a lithium-ion battery would therefore be desirable and can for instance be obtained by strain sensors due to increasing distension over the lifetime, causing e.g., strain on clamped battery packs [14]. Electric strain sensors however are only partially suited for the task, since precise measurement is only possible using bulky 3- or 4-wire connections and strong electric currents inside the battery system can impact the measurement.

One well-known optical method for strain and temperature measurement is the application of fiber Bragg gratings (FBG), which combine many advantages [15]. They are small and lightweight, and a large number of sensor elements can be cost-effectively integrated in a single glass fiber, thus reducing cabling complexity. They are mostly insensitive to electric and magnetic fields and able to withstand high temperatures of several hundred degrees Celsius [16,17].

FBG sensors have already been used for status monitoring of lithium-ion batteries by different research groups [18], since they are capable of delivering information beyond conventional BMS. They can serve as simple temperature sensors [19,20], outside and even inside of a cell [21,22], or as functionalized sensor elements that are sensitive to chemical substances and their optical properties [23,24]. Additionally, due to the small fiber diameter of 125 μm, FBG sensors can be integrated in many areas that are usually inaccessible for electric sensors e.g., in a battery system consisting of many densely stacked cells. FBG, applied as external strain sensors for monitoring lithium-ion cells with a flexible casing (also referred to as pouch cells), have been demonstrated several times already [25–27]. This observable strain variation during operation is of grea<sup>t</sup> interest for an improved state estimation because none of the aforementioned methods used in practice are capable of delivering information about the mechanical behavior of the cells, e.g., swelling due to aging effects.

For the readout of FBG sensors, several different interrogators exist [28]. Interferometric methods [29,30], tunable-filter-based interrogation methods [31] or methods with conversion from the wavelength domain to the time domain [32] are described in literature. All mentioned methods however are realized in discrete, bulky optical setups, built and adjusted by hand, even in the case of commercial devices and not robust enough, e.g., against vibrations, for reliable utilization in automotive applications.

Arrayed waveguide gratings (AWG), commonly used as (de)-multiplexing devices in telecommunications [33], are also well known as FBG interrogators for many years and their advantages, such as high number of output channels, wide bandwidth, precise wavelength detection, high-speed capability and low cost, have been reported [34]. Additionally, their compact, integrated design leads to a certain robustness that discrete systems cannot offer.

Most AWG are made of silicon and operate at wavelengths around 1550 nm [35]. In special applications—as it is the case with FBG-based battery monitoring at least today—the costs of manufacturing process setup as well as the cost-per-unit for silicon AWG is so high, that we could find no scientific or commercial implementation of AWG-based readout for fiber optic battery sensors despite the huge advantages. We therefore manufactured all-polymer AWG that are working at near IR center wavelength (850 nm), reducing manufacturing complexity, post-processing e ffort and resulting per-unit cost dramatically, as described in [36]. A further advantage of the unit working in the 850 nm region lies in lower cost of peripheral components like light sources and detectors.

Here, we present to our knowledge the first interrogation system based on an all-polymer AWG that is read-out by a CMOS linear image sensor and utilized as a readout-device for fiber optic battery sensors.
