Authors: Hanliang Zhu, Yue Zhang, Lan Wang, Jan Brodský, Imrich Gablech, Jianguo Feng, Qi-Long Yan, Shujie Yang, Luke P. Lee, Pavel Neuzil
Categories: Article, Electrical and electronic engineering
Source: Nature Communications
Authors: Hanliang Zhu, Yue Zhang, Lan Wang, Jan Brodský, Imrich Gablech, Jianguo Feng, Qi-Long Yan, Shujie Yang, Luke P. Lee, Pavel Neuzil
Calorimetry is crucial in biology, chemistry, physics, and pharmaceutical research, enabling the detection of heat changes at nanowatt and picowatt levels. However, traditional calorimetry systems are often limited by high costs and complex fabrication processes. Here, we reduce the cost and fabrication complexity of microcalorimeters by utilizing widely available flexible printed circuit manufacturing processes. This device achieves temperature and power resolutions of ≈ 6 μK and ≈ 654 pW in vacuum. Its feasibility is validated across a wide range of measurements, including salt crystallization, protein crystallization, and cellular metabolism. Our concept enhances the accessibility of microcalorimeters for high-resolution thermal analysis, which is challenging for conventional calorimeters.
Calorimeters are highly sensitive tools for direct heat measurement from physical changes, chemical reactions, and biological processes^1–3^. The development of calorimeters dates back to the end of the 18^th^ century when Lavoisier and Laplace invented the first calorimeter^4^. There was a boost in calorimetry development in the 20^th^ century due to the evolution of micromachining based on planar technology. The devices became much smaller with increased precision and sensitivity^5–8^. The classic structures of a micromachined microcalorimeter typically consisted of Si substrate, suspended SiO2, or low-stress SixNy membrane^9^ integrated with either a resistance temperature detector^10^, thermopiles, thermistors, or other temperature sensors. The thermally isolated miniaturized calorimeter chips exhibit a thermal time constant (τ) between ≈ 0.5 s to a few hundred seconds^5^.
Chip-based microcalorimeters have been extensively used for bioanalysis, providing precise measurements of heat changes during various biochemical processes. Their role in this field is complex and essential for understanding biological reactions, such as enzyme kinetics^11^, protein-ligand interactions^12^, and cellular metabolism^13^. Microcalorimeters enable researchers to investigate the thermodynamic properties of biomolecules by detecting small heat changes occurring in biological samples. This capability supports drug discovery and development^14,15^, as evaluating the binding affinities between drugs and their targets is critical. Microcalorimeters are categorized into two main types based on their structural design and fluid-handling closed and open systems^5^. Closed systems feature enclosed microfluidic channels that enable advanced functionalities, including fluid mixing, continuous sample injection, and encapsulation in controlled environments, such as a vacuum. However, the necessary microchannel-based sensor and the fluid within the system increase thermal mass and heat loss, limiting sensitivity and complicating the fabrication process due to the multistep requirements for the microfluidic structure, particularly in a suspended microcalorimeter^16^. In contrast, open systems simplify fluid handling by eliminating microfluidic channels, allowing samples to be directly placed onto the chip’s surface^17^. This design minimizes the chip’s thermal connection to the substrate, reducing heat loss. The sample is loaded into the microcalorimeter using either pipettes or nano dispensers. Although the open system lacks microchannels and continuously supplying fluids is challenging or possibly unfeasible, open systems are easy to be used for thermal analysis, such as measuring the enthalpy change of chemical reactions^17^, cell metabolism^18,19^, and molecular interactions^20–23^.
The microcalorimetry systems were constructed with an astonishing resolution in the tens of pW range. However, their fabrication process, which involves multiple photolithography steps, metal deposition, etching, and membrane release, is time-consuming and expensive^13,20,24^. A state-of-the-art microcalorimeter chip with a resolution of ≈ 28 pW was produced on a suspended SiN membrane using NiCr and CuNi thermopiles as temperature sensors^20^. It necessitates a complex fabrication process consisting of four-step photolithography and lift-off procedures to define the heater and sensor. Furthermore, it also includes Si substrate etching for membrane suspension, requiring an additional photolithography step. Creating flow-through microcalorimeters with microfluidic channels is even more involved, adding extra steps for channel formation^13^. Even commercially available equipment, such as the Flash DSC-1, continues to utilize costly Si-based chips, contributing to higher testing expenses^25,26^. Addressing these challenges ensures chip consistency and stability, enhancing the microcalorimeter’s performance and reliability by reducing the impact of intricate and costly microfabrication processes alongside inherent microfabrication costs.
We adopted flexible printed circuit board (FPCB) technology to fabricate a sensitive, reliable picocalorimeter, achieving a power resolution (PR) of (654 ± 67) pW (mean ± standard deviation, SD). This innovation enables the generation of extensive molecular thermodynamics data for monitoring the behavior of living cells in smaller sample volumes than traditional microcalorimeters, thus allowing higher accuracy and shorter response time. Our thermal simulation and optimized design demonstrate exceptional on-chip thermal isolation, achieving an impressive thermal conductance (G) (109.0 ± 8.2) µW·K^−1^ and (334.6 ± 32.9) µW·K^−1^, both (mean ± fitting error) in vacuum and air, respectively, constrained by available FPCB parameters from further improvement. Integrating the FPCB-based picocalorimeter into a low-noise measurement platform, we achieved a temperature resolution (TR) of ≈ 6 µK after calibrating it using light radiation and water droplet evaporation. Through experimental validation, we illustrated the effectiveness of a picocalorimeter by investigating the thermodynamics of protein crystallization within droplets and accurately monitoring metabolic heat dynamics in living cells. Our picocalorimeter’s sensitivity, simplicity, reliability, robustness, and affordability make it ideal for precisely quantifying biological and chemical reactions with a sub-nanowatt resolution, advancing molecular thermodynamics and quantitative biology at the single-cell level.
Our simple, low-cost picocalorimeter, with a power resolution of (3.68 ± 1.08) nW and (654 ± 67) pW (mean ± SD) in air and vacuum, respectively, can compete with more expensive published systems that have sub-nW range resolution^13,20,24^. The system also offers easy handling, requiring only a pipette for sample loading, whether it is a single sample droplet or a droplet covered with mineral oil, which forms a virtual reaction chamber (VRC)^27^ to prevent water evaporation. We extended the application of the VRC to operate in a vacuum, replacing the standard mineral oil used for polymerase chain reaction with low-saturated vapor pressure oil typically used for rotary vacuum pumps.
The heat (P) generated from a sample such as living cells is transferred to the surrounding environment and then measured by the temperature-sensing thermistor on the picocalorimeter (Fig. 1a). The heat balance equation describes the heat change (ΔT) in the picocalorimeter caused by P:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {C}_{{{\rm{P}}}}\frac{d\Delta T}{{dt}}+G\cdot \Delta T=P,
We selected a commercially available negative temperature coefficient (NTC) thermistor in a surface-mount device (SMD) type (Fig. 1e) with dimensions of ≈ (0.4 × 0.2 × 0.2) mm^3^ and a local temperature coefficient of resistance (TCR) of −4.95 % at 20 °C (Supplementary Fig. 1) (Supplementary Note 1). We designed the picocalorimeter with the size of ≈ (45 × 45) mm^2^ consisting of a hollow PI membrane, a Cu foil layer for electrical connection, and an SMD thermistor using JavaScript-based Nanolithography toolbox software^29,30^. The design was then sent to the PCB manufacturers, providing a simple and low-cost fabrication at 115 $ for 100 pieces (Fig. 1f). A complete PCB consisted of two FPCB picocalorimeters with soldered thermistors and the balancing thermistors for the Wheatstone bridge, prepared using the surface mounting technique. We selected four thermistors with nearly identical *R*~0~ values to minimize the Wheatstone bridge offset. We mounted these four thermistors at two one at the measuring, one at the reference calorimeters, and the other two thermistors at the rigid PCB. At the PCB, we added a few soldering pads to compensate for the Wheatstone bridge offset caused by differences in *R*~0~ values by adding a parallel or in-series resistor to the thermistor (Fig. 1g). We placed the complete PCB with two picocalorimeters into a three-layer thermally isolated measurement platform chamber, designed to minimize the temperature drift and capable of providing a vacuum environment (Fig. 1h). Temperature stability is essential for sub-nW and even nW resolution calorimetry because external factors such as airflow and additional light radiation are unavoidable. The chamber consisted of three the outer and middle layers were brass, and the inner layer was acrylonitrile-butadiene-styrene (ABS). The brass layers were equipped with an independent temperature control system. ### Simulation and optimization We built an analytical and numerical model with steady-state conditions to perform a thermal analysis of the FPCB picocalorimeter (Supplementary Fig. 2a). The numerical model is based on a three-dimensional (3D) realistic structure and was solved using the finite element method (FEM) in ANSYS software without the frame of the chip to simplify the simulation, as we could neglect its influence. The parameters of all materials for thermal simulation, such as thermal conductivity (λ), specific mass (ρ), thermal capacity (*C*~p~), and emissivity (ε), were listed in Supplementary Table 1. The *G* value of the system consists of conduction (*G*~cond~), radiation (*G*~rad~), and convection (*G*~conv~) (Supplementary Fig. 2b). The *G*~cond~ is calculated 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {G}_{{{\rm{cond}}}}=\left(4\cdot \frac{{w}_{{{\rm{PI}}}}\cdot {h}_{{{\rm{PI}}}}}{l}{{{\rm{\lambda }}}}_{{{\rm{PI}}}}+2\cdot \frac{{w}_{{{\rm{Cu}}}}\cdot {h}_{{{\rm{Cu}}}}}{l}{{{\rm{\lambda }}}}_{{{\rm{Cu}}}}\right)., $$\end{document}Gcond=4⋅wPI⋅hPIlλPI+2⋅wCu⋅hCulλCu.,where *w*~PI~, *h*~PI~, *l*, and λ~PI~ are the width, height, length, and thermal conductivity of supported PI legs, respectively; *w*~Cu~, *h*~Cu~, λ~Cu~ are the width, height, and thermal conductivity of Cu leads (Supplementary Fig. 2c) (Supplementary Note 2). The central test area on the picocalorimeter was a circular disc with a diameter of ≈ 3 mm by considering a maximum oil volume of ≈ 2 µL and a *CA* of 90° with a diameter of ≈ 2 mm. The width of the Cu and PI lines were set to 50 µm and 1 mm, respectively, both limited by the FPCB process. The heat loss caused by different radiations (*G*~rad~) from the picocalorimeter and the surrounding walls of the temperature-controlled chamber is defined by the Stefan-Boltzmann equation^31^:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {G}_{{{\rm{rad}}}}=\frac{{{{\upvarepsilon }}}\cdot A\cdot {{{\upsigma }}}\cdot ({T}_{1}^{4}-{T}_{2}^{4})}{{T}_{1}-{T}_{2}}, $$\end{document}Grad=ε⋅A⋅σ⋅(T14−T24)T1−T2,where ε, *A*, and *T*~1~ are emissivity, surface area, and average temperature of the radiation surface, respectively; *T*~2~ is the average temperature of the environment, and σ is the Stefan–Boltzmann constant. The heat loss in the picocalorimeter’s radiation was associated with *A* and ε. The non-uniform temperature distribution on the PI membrane complicated the calculation. We focused on the central testing area of PI with an ε value of ≈ 0.8 and a diameter of ≈ 3 mm, disregarding the thermistor and sample contributions. We performed simulations and optimizations on the key structural parameters of the picocalorimeter based on the FEM model in ANSYS. The *G* value was an essential indicator of the picocalorimeter calculated using the equation Δ*T* = *P*/*G*. A heat power of *P* = 7.8 μW was applied to the thermistor, resulting in the temperature distribution in the picocalorimeter and a maximum steady-state Δ*T*. The contribution of radiation and conduction to the *G* value in vacuum conditions was simulated and analyzed as a function of *l* of the leg. The length *l* of the leg increased from 5 mm to 25 mm, and the *G*~cond~ gradually decreased, while *G*~rad~ increased, leading to a decrease in the total *G* value. The *G* value reached an equilibrium having *G* value optimized at ≈ 106.3 μW·K^−1^ using *l* = 15 mm (Fig. 2a). We studied the influence of the *h*~Cu~ and *h*~PI~ parameters on *G* value using the FEM method. However, they were limited by the standard process of FPCB to ≈ 12 µm and ≈ 25 µm, respectively (Fig. 2b). In conclusion, we identified the optimal design we set the PI leg thickness to 12.5 µm with a *w*~PI~ of 1 mm; the Cu layer thickness to 12 µm with a width of 50 µm; and we designed the *l* of 15 mm. The simulation results suggest that we can further optimize the *G* value of the chip by reducing the thickness of the Cu leads down to the μm or even nm-level with the future potential development of FPCB technology.Fig. 2Simulation, optimization, and fabrication stages of the FPCB picocalorimeter.**a** Graph illustrating the relationship between thermal conductance (*G*~total~) and the leg length (*l*), showing the contribution of conduction (*G*~cond~) and radiation (*G*~rad~) to the overall thermal conductance in a vacuum environment. **b** Analytical model representing the impact of Cu thickness on *G*~total~, plotted against PI thickness, to optimize the picocalorimeter’s design for thermal isolation. **c** Finite element analysis (FEM) displaying temperature distribution along the picocalorimeter’s PI legs, indicating the thermal impact on the structure with a zoomed-in view for detailed analysis. **d** FEM heat distribution simulation across the picocalorimeter, with a top and cross-sectional view highlighting the temperature gradient (∆*T*) due to a *P*~heat~ value of ≈ 7.8 µW in the center. **e** Stepwise schematic of the FPCB-based picocalorimeter fabrication (i) Starting with the original PI plate with rolled Cu, (ii) followed by lithography and Cu etching, (iii) laser cutting for precise shaping, and (iv) reflow soldering of the surface mounted device (SMD) thermistor, culminating in the assembly of the complete PCB with soldered PI calorimeters. We then applied the same *P*~heat~ value of ≈ 7.8 µW on the thermistor for temperature distribution of the optimized FPCB picocalorimeter. The maximum temperature rise in the testing area of the picocalorimeter was ≈ 74 mK, mainly concentrated in the central region (Fig. 2c). We then introduced a sample droplet of 400 nL into the model. We applied an identical *P*~heat~ value to the sample. The simulation result of the sample/picocalorimeter cross-section showed the temperature distribution variation from a minimum Δ*T* of ≈ 55 mK next to the sensor to a maximum Δ*T* of ≈ 59 mK in the sample, showing that the sensor has an offset of 7% concerning the temperature at the top of the sample (Fig. 2d). This indicates that the environmental heat exchange influences the thermal distribution of the chip system, and further optimization of the *G* value could reduce this deviation. Using Eq. (2) and Eq. (3), the calculated values of *G*~cond~ and *G*~rad~ were ≈ 35.5 μW ⋅ K^−1^ and ≈ 64.8 μW ⋅ K^−1^, respectively, leading to a total *G* value of ≈ 100.4 μW·K^−1^, which is consistent with the FEM results. Furthermore, we performed thermal simulations under ambient conditions by introducing convection heat transfer boundary conditions into the model (Supplementary Fig. 2d). We analyzed the effects of conduction, radiation, and convection on heat loss in the picocalorimeter under ambient and vacuum conditions. In an ambient environment, the heat loss of the picocalorimeter primarily comes from air convection, accounting for ≈ 63.3%, more than the heat loss from conduction and radiation. In contrast, the *G*~rad~ value becomes more critical than conduction heat loss in a vacuum (Supplementary Fig. 2e-f) due to the extensive testing area with a diameter of up to 3 mm. The fabrication started with photolithography for patterning a ≈ 12 µm thick Cu foils thermally laminated to a ≈ 25 µm thick PI substrate (Fig. 2e–i). This photolithography utilized a dry film photoresist (DFP) exposure to UV light via mask. The exposed areas of the photoresist were crosslinked, and a chemical developer dissolved the unexposed regions. The unprotected copper areas were etched using an etchant solution, leaving the Cu pattern on the substrate (Fig. 2e–ii), and the DFP was removed with a photoresist stripping solution. The last step for the single-layer FPCB was the electroless plating of an Au layer at the Cu to prevent its surface oxidation and enable the soldering of wires for electrical connection. An additional laser cutting formed the PI legs for thermal isolation of the picocalorimeter center area (Fig. 2e–iii). Finally, the soldering pads were covered with solder paste, and two SMD NTC thermistors were placed on two FPCB chips, one at each. The two FPCB chips were subsequently mounted on a rigid PCB using a reflow soldering technique (Fig. 2e–iv) (see “Methods”). ### Characterization of the system Temperature sensors at two picocalorimeters were connected into a Wheatstone bridge with two outputs, *V*~1~ and *V*~2~, subtracted from each other and amplified by a differential preamplifier (Fig. 3a). A lock-in amplifier further processed the output signal (*V*~1−~*V*~2~) to increase the signal-to-noise ratio. Then, we used an oscilloscope to monitor and store its output voltage (Supplementary Fig. 3a). We used the self-heating method^32^ by generating Joule heat on the thermistor to measure the system’s thermal parameters, including *G*, *C*~p~, and τ, both in vacuum and air (Supplementary Note 3).Fig. 3Characterization and calibration of the FPCB picocalorimeter system.**a** The circuit diagram of the test platform for the picocalorimeter. The printed circuit board (PCB) with soldered two picocalorimeters, one for a sample and one for a reference was balanced with two identical thermistors forming an AC-powered Wheatstone bridge with an amplitude of *V*~B~. The bridge outputs were subtracted from each other using a differential amplifier (*V*~1~ − *V*~2~), and a lock-in amplifier processed the output differential signal. **b** Thermal conductance (*G*) measurement is based on a self-heating picocalorimeter by dissipated Joule heat as a function of *V*~B~. The *G* in a vacuum with the pressure of ≈ 16 mPa and air was (109.0 ± 8.2) µW·K^−1^ and (334.6 ± 32.9) µW·K^−1^, both (mean ± fitting error). **c** The system’s τ value was measured using light pulse illumination with the nominal wavelength of 490 nm as a heat source, giving the value of τ of (1.106 ± 0.001) s and (4.057 ± 0.001) s in air and vacuum, respectively, both (mean ± fitting error). **d** The picocalorimeter was placed into the temperature-stabilized vacuum chamber to evaluate its sensitivity and noise level. The system’s sensitivity was ≈ 24.75 mV·K^−1^ with V~B~ set to 2 V using a thermistor with a temperature coefficient of resistance (TCR) of ≈− 4.95% K^−1^ at 20 °C. The temperature noise was ≈ 11 µK and ≈ 6 µK in air and vacuum, respectively, resulting in the corresponding power resolution of (3.68 ± 1.68) nW and (654 ± 67) pW, both (mean ± standard deviation). **e** The plot of the picocalorimeter’s *G* value as a function of ambient pressure (*p*) in the logarithmic scale demonstrates the system’s operational window. It indicates the vapor pressure threshold of ≈ 10^−5^ Pa of the oil used in the system. (**f**) Summarization of published work, giving each power resolution and two key parameters, thermal conductance (*G*) and temperature (*T*) resolution, where the limit of detection (LOD) is defined as *P* = *G*·*T*. We measured the Δ*T* value caused by Joule heating due to the applied *V*~B~ at the picocalorimeter, obtaining *G* value of (109.0 ± 8.2) µW·K^−1^ and (334.6 ± 32.9) µW·K^−1^, both (mean ± fitting error) in vacuum and air, respectively, calculated from the slope of the Δ*T* as a function of *P* (Fig. 3b). These obtained results (in vacuum) were consistent with the *G* value determined by FEM of ≈ 100.4 μW·K^−1^ and calculated by Eqs. (2) and (3) of ≈ 106.3 μW·K^−1^. The picocalorimeter’s τ value was evaluated by increasing the Δ*T* using a long duration of ≈ 43 s of light radiation as a heat source to ensure the picocalorimeter reached its thermal equilibrium^33^. We then conducted data fitting of system Δ*T* using an exponential curve of the first order by Eq. (S11) to extract τ value as (4.057 ± 0.001) s and (1.106 ± 0.001) s, both (mean ± fitting error) in vacuum and ambient pressure, respectively (Fig. 3c), achieving a coefficient of determination *R*^2^ of 0.999. We extracted the τ value during system passive cooling, getting practically identical values. We calculated the *C*~p~ value of the FPCB picocalorimeter using an equation *C*~p~ = *G*·τ, obtaining the value of (442.2 ± 33.3) µJ·K^−1^ and (368.0 ± 36.4) µJ·K^−1^, both (mean ± SD) in vacuum and air, respectively. The resolution of the picocalorimeter system is primarily determined by two the thermal isolation of its membrane, indicated by its *G* value, and the inherent temperature noise within the system. Electrical noise from the testing circuit and environmental heat fluctuations also impact the resolution. Our picocalorimeter was placed in a three-layer-based independently temperature-controlled chamber, achieving a temperature control precision of ± 300 µK (Supplementary Fig. 3b–d). Additionally, adding another layer of low thermal conductivity insulation further improved the temperature stability of the test chamber. Generally, the picocalorimeter’s short-term and long-term noise significantly affects the system’s performance. Short-term noise indicates immediate stability, which impacts the accuracy of measurements during dynamic processes such as protein folding or chemical reactions. We determined the noise level by measuring temperature variations over 100 s to quantify short-term noise. During testing, the voltage bias (*V*~B~), sensitivity (*S*), and gain (*B*) of lock-in amplifier values were set to 2 V, 50 mV, and 200, respectively. We measured the system SD of the voltage noise at ≈ 155 nV and ≈ 263 nV in a vacuum and air, corresponding to temperature noise of ≈ 6 µK and ≈ 11 µK, respectively (Fig. 3d). As a result, our picocalorimeter achieved a power detection limit (*P*~R~) in a vacuum and air of (654 ± 67) pW and (3.68 ± 1.08) nW calculated using *P*~R~ = *G* × Δ*T*. Long-term noise represents the system’s long-term stability. It significantly impacts experiments requiring extended monitoring periods, such as cellular metabolism or biological reactions under changing environmental conditions. The system had long-time noise over 10 h of ≈ 36 µK in the air (Supplementary Fig. 3e in Supplementary Note 4). The *G* value of a picocalorimeter can be expressed as the sum of the following three *G* = *G*~cond~ + *G*~rad~ + *G*~conv~, as discussed above, primarily contingent upon the structure and material used. Conversely, *G*~conv~ is influenced by ambient pressure (*p*~A~), and this relationship has been well-documented^34^. The picocalorimeter’s geometry and proximity to the surrounding environment determine the critical pressure point. This pressure can be simulated using appropriate finite element analysis techniques^9^ or directly measured. The measurement involves determining the *G* value as a function of *p*~A~. We utilized the lock-in amplification method with a *V*~B~ set to 2 V RMS with ≈ 7.8 µW of Joule heat dissipation at the picocalorimeter membrane, varying *p*~A~ as a parameter. The temperature of the picocalorimeter membrane was heated up to a maximum ∆*T* of ≈ 71 mK corresponding to the measured *G* value of (109.0 ± 8.2) µW·K^−1^ and (334.6 ± 32.9) µW·K^−1^, both (mean ± fitting error) in vacuum and air, respectively, with the lowest *p*~A~ of ≈ 14 mPa. We applied an ≈ 80 s long voltage pulse with a set bias of 2 V, which generated a dissipated power of ≈ 7.8 µW while changing the *p*~A~. We recorded the lock-in amplifier voltage output (∆*V*~L~) and transferred it to ∆*T* to determine the *G* value as a function against *p*~A~, which yielded a characteristic sigmoidal curve (Fig. 3e). In the previous research, we used mineral oil to prevent the evaporation of sample droplets; however, it vaporized in a vacuum with a saturated vapor pressure of ≈ 0.5 Pa. In response, we explored an oil for a rotary vacuum pump with a saturated vapor pressure of ≈ 10 µPa, successfully substituting the mineral oil and enabling testing in high vacuum conditions. Our investigation unveiled a critical operational pressure threshold of ≈ 1 Pa or lower for sustaining optimal system performance. The fundamental parameters that define each calorimetric system, *P*~R~, *G*, and *T*~R~, are closely related yet influence the system’s performance independently. This connection and the individual impact of each parameter become evident when comparing various published microcalorimeter systems^9,13,16–18,20,24,35,36^ (Fig. 3f), as listed in Table 1. Achieving extraordinary advancements, the *P*~R~ of microcalorimeters was as low as ≈ 28 pW in open space systems, accompanied by a *G* value of ≈ 100 µW·K^−1^ on SiN membranes and an impressive *T*~R~ of ≈ 0.28 µK due to optimized thermopiles^20^. The best thermal isolation for a microcalorimeter system was achieved using a suspended parylene membrane^13^ having a remarkably low *G* value of ≈ 2.5 µW·K^−1^. On the contrary, the *T*~R~ is contingent upon the sensor type and the environmental conditions, now exhibiting a spectrum from ≈ 0.28 µK to ≈ 3300 µK. These two critical parameters, *G* and *T*~R~, exhibit no direct correlation.Table 1Comparison with state-of-the-art technology*P*~R~(nW)G(μW/K)*T*~R~(μK)τ(s)Volume(nL)(mV·K^−1^)FabricationRef0.7109.06.31.120–100024.75One side FPCThis work0.0281000.281.1 − 2.2500 − 100013.345-step photolithography, and multistep deposition, etching^20^0.22.58010/0.2297-step photolithography, and multistep deposition, etching^13^0.27271010045130Four-side Au deposition and etching^24^4.2162701.33.50.118-step photolithography and multistep deposition, etching^16^11.5633000.50.490.01654-step photolithography and multistep deposition, etching^9^14283.8491.40.50.081-step photolithography and Au deposition, etching^18^1729005.70.716128015.64-step photolithography and multistep deposition, etching^35^22500441.1501.45Combined with Xensor (XEN-NCM9924)^17^70701000850,0000.08Combined with infrared sensor (S25)^15^ Our picocalorimeter, fabricated using conventional and simple FPCB technology, demonstrated exceptional performance. It has excellent parameters, such as a *G* value of (109.0 ± 8.2) µW·K^−1^ (mean ± fitting error) in vacuum, and a *T*~R~ value of ≈ 6 µK, achieving a power resolution in sub-nW. Therefore, its superior performance offers researchers a more accessible and convenient method for conducting calorimetry and microcalorimetry. ### Measurement using FPCB picocalorimeter We developed this FPCB picocalorimeter for precise quantitative biological, chemical, and life science research. Measurements involving five different samples were system’s accuracy evaluation by measuring latent heat of water evaporation, response to prolonged light pulses irradiation, thermodynamics of protein crystallization caused by evaporation, energy release from the death of *Paramecium caudatum* (*P. caudatum*) during evaporation, and stability of droplets in vacuum. #### Evaluation of the system using water evaporation The accuracy of the picocalorimeter system was evaluated through measurements of the heat change during water evaporation, compared against the known latent heat of water evaporation of 2453.5 kJ·kg^−1^ at 20 °C. The latent heat is a stable thermodynamic constant that is a reliable reference for calibrating the calorimetry systems^37^. This measurement takes advantage of the phase change characteristics of water, particularly the transition from liquid to vapor, which involves significant latent heat exchange. We demonstrated that the droplet evaporation process on the picocalorimeter surface follows a stick–slide mode shown in previous work^38,39^. The evaporation rate correlates with the perimeter of the solid-liquid contact line^40–42^, initially remaining constant before decreasing rapidly near the end of evaporation. We also developed a script for the MATLAB environment to calculate the *P* value based on Eq. (5) with variable *C*~p~ and *G* values during evaporation (Fig. 4a)^38^.Fig. 4System accuracy verification by measuring droplet evaporation heat.**a** The heat flow calculation of the system is based on a heat balance equation. **b** The power required to evaporate the water droplet was analyzed using a heat balance equation conducted by a MATLAB script. **c** The latent heat of water evaporation at 20 °C was calculated to be ≈ 2600 kJ·kg^−1^. The heat generation standard curve of the droplet evaporation with different volumes from ≈ 200 nL to ≈ 600 nL shows the precision of the pipetting and power measurements. **d** Evaluation of the sensitivity of the picocalorimeter system to fast heat changes by utilizing short pulses. The pulse width was set from ≈ 5 ms to ≈ 25 ms with the step of ≈ 5 ms. **e** The temperature increased by an incident radiation caused by the short light pulse was recalculated to power using a MATLAB program. We first waited several minutes to stabilize the system before pipetting a droplet of ≈ 400 nL of deionized (DI) water onto the chip surface to test for natural evaporation. The droplet entirely evaporated over ≈ 551 s. The results revealed an almost linear evaporation behavior^43^, with a ≈ 43.6 nL·min^−1^ rate consistent with the known proportionality between evaporation rate and droplet perimeter. The Δ*T* during droplet evaporation was calculated from the measured Δ*V* using the equation Δ*T* = 4Δ*V*/(α*V*~B~). The Δ*P* was converted using the script in MATLAB environment and then integrated the *P* over *t* to obtain the total heat absorbed during evaporation of (1010.87 ± 33.44) mJ (mean ± SD). Multiple water evaporation experiments were conducted using droplet volumes set to 200 nL, 300 nL, 400 nL, 500 nL, and 600 nL to validate the method further (Fig. 4b). The evaporation of each volume was tested three times to ensure reproducibility. Integration of *P* over *t*, followed by linear regression of the heat as a function of water mass, produced a latent heat of (2600.0 ± 51.5) kJ·kg^−1^ (mean ± fitting error) (Fig. 4c). This value deviated by ≈ 5.97 % from the theoretical latent heat of evaporation at 20 °C of 2453.5 kJ·kg^−1^, confirming the accuracy of the system’s measurements. #### Transient response measurement The *P* calculation method, as described in Eq. (5) and based on the heat balance equation in the MATLAB program, effectively accounts for the transient heat capacity effect of the system through *C*~p~·d*T*/d*t*. Furthermore, the τ value describing the system’s response was extracted from the data as a function of *C*~p~/*G*. This approach offers greater accuracy in handling dynamic heat changes than methods that rely solely on the steady-state formula^13,24^ *P* = *G* × *T*, particularly when the τ value exceeds 10 s. The system can detect rapid heat changes in several orders of magnitude smaller than its τ of (1.106 ± 0.001) s (mean ± fitting error) in air, significantly enhancing its capability to capture transient heat pulse behavior. This improvement enhanced the system’s performance in rapid heat change testing, even in systems with greater τ values. We performed transient heat measurements by applying short light pulses ranging between 5 ms and 25 ms on the picocalorimeter to capture the response of instantaneous heat changes. Direct emission LEDs typically have a turn-on time in single-digit nanoseconds and turn-off times in the tens of nanoseconds. We investigated the performance of the picocalorimeter system by using short pulses of light generated from LED emission. The LED was powered by a signal generator with set pulse durations of 5 ms, 10 ms, 15 ms, 20 ms, and 25 ms. The *V*~*B*~ of the system was set to 15.82 V, and *S* of the lock-in amplifier and *B* of AD8221 were set to 10 mV and 5.1, respectively, resulting in a system sensitivity of ≈ 195.8 mV·K^−1^. We then evaluated the sensitivity of the picocalorimeter system by utilizing short voltage pulses, causing rapid changes in membrane temperature. The resulting heat signals were captured and analyzed, allowing us to determine the system’s ability to detect and measure these fast heat changes accurately (Fig. 4d). The heating time of 5 ms caused the sensor’s temperature to increase by ≈ 1.4 mK and ≈ 25 ms up to ≈ 7.2 mK. The *V*~*L*~ data was then processed in the MATLAB program to convert *V*~*L*~ to *P* (Fig. 4e). The obtained results confirm the effectiveness of our system in employing picocalorimetry for applications requiring precise and rapid detection in thermodynamic studies. #### Evaporation crystallization of protein Investigating the crystallization process of protein solutions during evaporation^44^ is a significant area in structural biology and biophysics. Understanding how proteins form their 3D structures by crystallization is vital for drug discovery and drug development^45^, and it also provides essential information about their behavior^46,47^. However, optimizing crystallization conditions remains challenging due to its complex, trial-and-error nature. Enhancing reproducibility and optimization in this process is essential. As water from the protein solution evaporates, the remaining proteins often form characteristic patterns known as *coffee rings*^48,49^, and the crystals that emerge can vary in size, shape, and density depending on factors like salt concentration^50^ and protein folding/unfolding status^51^. Our experimental study employs picocalorimetry and microscopy techniques to investigate pattern formation in dried bovine serum albumin (BSA) droplets in folded and unfolded conformations within a KCl solution. As solvent molecules gradually depart during evaporation, solute concentration increases, potentially initiating nucleation and subsequent protein crystal growth. This process is driven by alterations in solute-solvent interactions and considerations of entropy. We placed a ≈ 400 nL BSA solution at a concentration of ≈ 50 mg·mL^−1^, containing varying KCl concentrations, onto a hydrophobic-coated cover glass for natural drying observations (Fig. 5a). The BSA at 50 mg·mL^−1^ without KCl presence exhibited a typical *coffee ring* pattern. As the KCl concentration increased to ≈ 62.5 mM, a mesh structure formed within the *coffee ring*, marked by the ring edge cracks^52^. An interesting dendritic branch structure emerged at concentration of ≈ 0.25 M. Higher KCl concentrations significantly affected the patterns, evident in crystal clusters showcasing intensified salt crystallization. Concurrently, we monitored the pressure change during BSA solution evaporation with KCl (Fig. 5b). The evaporation properties remained consistent across all droplets until crystallization commenced. The introduction of KCl influenced crystallization time, mode, and final crystal formation, which is particularly evident in higher concentrations.Fig. 5Comprehensive analysis and application of the picocalorimeter in biological studies.**a** A series of images depicting crystallization patterns in dried droplets of bovine serum albumin (BSA) at various KCl concentrations illustrates the effect of salt on BSA crystallization. **b** The graph demonstrates real-time monitoring of power dissipation during droplet evaporation, with thermal curves revealing distinct crystallization behaviors attributable to different salt concentrations. **c** A combined thermal and microscopic analysis from a single test highlights the crystallization endpoint, featuring a rapid heat release of ≈ 41.02 µJ within ≈ 0.5 s. **d** Time-lapse photography captures the life cycle and eventual demise of *P. caudatum* due to water evaporation. **e** The images showing the process of measuring metabolic heat for varying numbers of *P. caudatum* within ≈ 400 nL droplets illustrate the calorimetric changes resulting from water evaporation and organism activity, with baseline calibration against a droplet containing only water. **f** Dissipated energy as a function of time is presented with the number of *P. caudatum* as a parameter. We obtained the energy by integrating the data shown in (**e**). Extracted values of total dissipated energy as a function of the number of *P. caudatum* indicate that their metabolic heat was linearly proportional to their quantity in the sample. Subsequently, we recorded a video during the evaporation of a ≈ 50 mg·mL^−1^ BSA solution with ≈ 0.25 M solution of KCl. BSA and KCl crystallized at ≈ 251.9 s, resulting in a sudden ≈ 0.6 s temperature increase. After this, the evaporation rate surged again, cooling the chip due to the dendritic branch structures on the surface, which increased the evaporation perimeter. We determined the energy released during crystallization to be ≈ 41.02 µJ (Fig. 5c) by conducting linear fitting and extrapolation around *t* ≈ 251.9 s as demonstrated^39^. The local minima at the curve at *t* ≈ 252.3 s occurred when the crystal seeds were formed, and the crystals started to grow. Then, the exothermic crystallization process seems to slow down, allowing the temperature to decrease further. Once the water evaporation rate slowed due to the lack of water in the solution, the sample *T* increased till it reached the original equilibrium. The BSA-KCl crystals formed from the droplet’s edge and expanded toward the center. #### Monitoring of *P. caudatum* death during evaporation Research on the drying of biofluid droplets containing living cells, such as pathogenic viruses, bacteria, algal dispersions, and multicellular microorganisms, has garnered significant attention^53–55^. Our study utilized *P. caudatum* as a model organism to demonstrate the capabilities of our picocalorimeter in monitoring the drying process of biofluid droplets. We investigated the thermodynamic characteristics of their physiological states and the processes leading to cell death by tracking the heat changes during the evaporation of droplets containing *P. caudatum*. This approach enables a deeper understanding of the metabolic activity, dehydration response, and death kinetics of living cells under controlled drying conditions. We employed a precise pipette to transfer ≈ 400 nL of the culture containing *P. caudatum* cells onto a picocalorimeter immediately after microscopic observation to investigate the impact of water evaporation on the protist*. P. caudatum*, reliant on its predominantly water-based cellular structure, showed increased vulnerability to water loss. This vulnerability was confirmed by significant disruptions in cellular functionality, including the contraction and greater fragility of cell membranes, potentially leading to ruptures and subsequent leakage of cellular components. A single *P. caudatum* cell, measuring ≈ 300 µm in length and a maximum diameter of ≈ 50 µm, accounted for ≈ 0.59 nL, making it significantly smaller than the ≈ 400 nL droplet initially used. Consequently, observable heat generation from the *P. caudatum* cells (*Q*~P.caudatum~) only became evident as the droplet evaporated to a specific volume. The droplet evaporated with its contact line anchored on a PI membrane, maintaining a constant perimeter and, thus, a consistent evaporation rate (Fig. 5d). The reduction in droplet height eventually restricted the movement of *P. caudatum* at ≈ 371 s, followed by further shrinkage of the water contact line, ultimately resulting in the demise of *P. caudatum* within the next 23 s. We monitored Δ*T* during water evaporation from droplets with a volume of ≈ 400 nL containing *P. caudatum* in a culture media and a sample with no *P. caudatum*. Temperature as a function of *t* of a droplet with no of *P. caudatum* exhibited a conventional shape of water evaporation from a droplet corresponding water evaporation rate influenced by droplet perimeter^39,41^ (Supplementary Fig. 4a, b). The total average *Q* value was (806.2 ± 11.3) mJ (mean ± SD from three measurements). We observed significant thermal perturbations due to the metabolic activity of the organisms once the droplets with *P. caudatum* were smaller due to the substantial amount of water that had evaporated (Supplementary Fig. 4e–f). The black lines represented the Δ*T* when a single (Supplementary Fig. 4e), two (Supplementary Fig. 4f), three (Supplementary Fig. 4g), four (Supplementary Fig. 4h), and five (Supplementary Fig. 4i) of *P. caudatum* were present with a red line representing the sample without *P. caudatum*. The metabolic activity of *P. caudatum* during the water evaporation stage resulted in additional heat release exhibiting deviation from the baseline. Once the organisms’ vital functions ceased, the Δ*T* returned to 0 K at a rate consistent with the baseline, signaling the death of the microorganisms. We used Eq. (5) to calculate the *P* value as a function of *t* with the number of *P. caudatum* as a parameter. We subtracted the baseline, getting Δ*P* value caused by *P. caudatum* metabolic activities (Fig. 5e). The increase in Δ*P* value was directly proportional to the number of organisms, indicating that metabolic heat release scales with the number of *P. caudatum*. The highest power output associated with individual *P. caudatum* cell deaths was ≈ 202.5 µW, temporarily affecting the temperature of the picocalorimeter. We integrated the Δ*P* value over *t*, getting the *Q* value as a function of a number of *P. caudatum*. Subsequently, we performed a linear regression, getting the *Q* value for a single event of *P. caudatum* death (*Q*~P. caudatum~) as (2.58 ± 0.28) µJ (mean ± fitting error) (Fig. 5f). The different Δ*T* as a function of *t* curves exhibited slight variations from each other, likely due to individual differences in the organisms, their death processes, and potential variations in the micro-environment of the droplets. This variability suggests that repeat experiments, even with the same number of organisms, may show some differences in the Δ*T* (and thus Δ*P* and *Q*) as a function of *t* curves (Supplementary Note 5). #### Open-space picocalorimeter in vacuum The characterization of the picocalorimeter both in ambient conditions and in vacuum indicated that vacuum conditions significantly enhance system performance. The results showed that in a vacuum environment, the chip’s *G* value and temperature noise decreased to one-third and one-half, respectively, twice, improving the system’s resolution about six times. Previously, we used mineral oil with a saturated vapor pressure of ≈ 0.5 Pa to form the outer shell of the VRC, effectively preventing water evaporation from the sample in an ambient environment. However, when the surrounding pressure drops to or below ≈ 0.5 Pa, the oil evaporates significantly, altering the VRC heat capacity and affecting picocalorimeter measurements. Here, we proposed replacing mineral oil with vacuum oil, which has a much lower saturated vapor pressure of ≈ 10 µPa, to eliminate the issue of oil evaporation in a vacuum environment. We verified the stability of this VRC with vacuum oil by testing its *C*~p~ value while exposing the picocalorimeter with a VRC made of water and vacuum oil to a vacuum environment for ≈ 180 min. The *C*~p~ value of the sample remained stable (Supplementary Fig. 5a, b), demonstrating the potential of the open-space FPCB picocalorimeter for operation in vacuum conditions (Supplementary Note 6). This method is particularly beneficial for long-term research or experiments requiring precise measurements under vacuum conditions, which have six times better sensitivity than in an ambient environment. The system’s maximum resolution under vacuum reached ≈ 654 pW, calculated by multiplying the noise value by the *G* value to define the system’s resolution limit. ## Discussion This study introduces a low-cost, highly reliable, and efficient picocalorimeter system fabricated using FPCB technology, making it accessible to laboratories with limited resources. The system’s simplicity in design and manufacturing, combined with the integration of off-the-shelf thermistors, allows for quick fabrication and direct use without complex calibration procedures. The cost of producing each chip is remarkably low, ≈ 1$, enabling widespread adoption in various research settings. The low fabrication cost makes the picocalorimeter disposable, eliminating sample cross-contamination and chemical degradation of materials during biological studies. A vendor can even coat a picocalorimeter with fluorosilane, making it convenient for users. This picocalorimeter system provides significant real-time advancements in monitoring heat changes in chemical and biological processes. Its applications are diversed, as demonstrated by experiments in water evaporation, protein crystallization, cellular metabolism, and more. This picocalorimeter can also be used for material studies, such as monitoring hydrogel properties, drying, and other physical changes^56,57^. The system’s flexibility, affordability, and efficiency allow high-performance calorimetry to be conducted in laboratories that may not have access to traditional, more expensive equipment. Microcalorimetry has long been recognized for its potential in studying protein crystallization, yet optimizing crystallization conditions remains challenging due to the trial-and-error nature of traditional methods. While differential scanning calorimetry (DSC) provides valuable information, its limitations in time resolution hinder real-time monitoring of rapid transitions that occur during crystallization. Our system addresses these limitations, offering an unparalleled ability to measure heat flow during crystallization with high time resolution. The real-time heat release during the crystallization of BSA-KCl droplets, recorded at ≈ 41.02 µJ, along with the subsequent formation of dendritic structures, underscores the system’s sensitivity and potential to deliver valuable thermodynamic data^52,53,58^. We gained insights into the thermodynamic and morphological changes that occur during crystallization by combining calorimetry with microscopy. This integration enhances our understanding of crystallization kinetics, providing a more comprehensive view of solute concentration, phase transitions, and protein structure formation. Furthermore, the ability to conduct experiments under varying conditions, including in a vacuum, allows for phase transition studies under controlled environmental parameters. This capability is particularly advantageous in structural biology and drug development, where precise control of crystallization conditions is essential. In addition to its role in protein crystallization, our picocalorimeter has significant implications for studying biological systems, particularly cellular metabolism. The analysis of *P. caudatum* metabolism during droplet evaporation demonstrated the system’s capability to monitor heat changes linked to cellular activity. The calculated metabolic heat release of (2.58 ± 0.28) µJ per organism provides insight into the thermodynamics of cellular processes during dehydration and cell death^58–61^. Tracking these metabolic processes in real-time offers a label-free method to study cell behavior. The detection of heat release during the cessation of metabolic functions and the eventual death of the organisms highlights the potential of this picocalorimeter system to explore complex biological phenomena. This high-resolution approach enables precise thermodynamic analysis of cell behavior in response to environmental stressors, such as dehydration, a common challenge in biological research. Furthermore, the system’s versatility, demonstrated through its application to other biological organisms, including bacteria and viruses, presents exciting opportunities for monitoring metabolic or inactivation processes in various biological systems^62^. The FPCB-based picocalorimeter system developed in this study shows significant advantages in cost-effectiveness and accessibility. The PI substrate, recognized for its chemical stability, was utilized throughout the experiments without any observed changes in parameters. However, long-term chemical stability and performance during extended observations, particularly in biological studies, require further investigation. Environmental factors, such as temperature fluctuations and mechanical vibrations, which could affect long-term calorimetric measurements, need to be studied and addressed to enhance the system’s robustness and accuracy. Despite its strengths, the system has inherent limitations. Operating exclusively in a constant temperature environment, it lacks integrated heating or cooling features, which limits its ability to support DSC or studies at extreme temperatures. Such functionalities are vital for applications involving rapid temperature changes or high-temperature operations. The current system’s operating temperature is constrained by the glass transition temperature of the solder alloy, ≈ 180 °C. Substituting the solder alloy with higher-temperature variants could extend this limit, although the polyimide substrate’s inherent properties would likely cap the operational range between 300 °C and 400 °C. Applications demanding higher thermal stability using alternative substrates, such as polybenzimidazole^63^, could be explored. Efforts are underway to upgrade the system to include isothermal calorimetry capabilities through an integrated temperature counterbalancing heater. This enhancement would enable the picocalorimeter to maintain a stable temperature by dissipating Joule heat, allowing for elevated temperature studies and facilitating real-time DSC experiments. Research also focused on enhancing the calorimeter’s resolution and thermal sensitivity. Exploring alternative temperature sensors, such as PN junctions^64^, sensor arrays, and materials with advanced thermal properties, along with sophisticated readout electronic circuits^65^, could significantly boost the device’s performance. These advancements would facilitate precise studies of rapid thermal transitions, including protein-ligand interactions and detailed metabolic profiling. Integrating microfluidic channels with the FPCB picocalorimeter similar to the one used for mass spectrometry/high-performance liquid chromatograpy^66^ could significantly enhance its functionality by allowing researchers to manage the flow of fluids or gases into the system. This represents a crucial direction for future development. Such integration would facilitate dynamic experiments under precisely controlled conditions, such as regulated chemical reactions or enzyme kinetics. Like the configurations used in mass spectrometry or high-performance liquid chromatography, this enhancement would improve the system’s capabilities, particularly in analyzing protein-ligand interactions and other biochemical processes. Further work is planned to develop a standalone system that incorporates computer numerical control-machined parts, enhancing system robustness and integration while ensuring cost-effectiveness. While prototypes are still in development, discussions about balancing performance and affordability emphasize the potential for this technology to make high-performance calorimetry accessible to a wider scientific audience. Additionally, optimizing packaging and enhancing the stability of droplet-based sample loading is essential to ensure consistent and reliable results. The proposed advancements aim to tackle existing limitations, broaden the device’s applicability across various scientific fields in biology, chemistry, or materials science, and uphold its affordability and simplicity. This study introduces a cost-effective, user-friendly open-space picocalorimeter utilizing FPCB technology. We characterized the system, achieving high power and temperature resolutions of ≈ 6 µK and ≈ 654 pW in both air and vacuum. The FPCB picocalorimeter was employed to analyze significant physical phenomena, such as salt and protein crystallization, as well as cellular metabolism. We measured the real-time heat release during the crystallization of BSA-KCl droplets to be ≈ 41.02 µJ, and the metabolic heat release from one *P. caudatum* was (2.58 ± 0.28) µJ (mean ± fitting error). These experiments demonstrated that our picocalorimeter can be used by researchers across various fields, including structural biology, cellular metabolism, and materials science. The system’s capability to monitor real-time thermodynamic changes under diverse experimental conditions, including in a vacuum, further enhances its versatility. As the technology develops, advancements in sensor resolution and system integration will broaden its applications, allowing researchers to investigate increasingly complex systems and processes. This system could transform how calorimetric measurements are conducted in research, making it more accessible to a wider range of laboratories and researchers. ## Methods ### Layout design and fabrication of the FPCB picocalorimeter We generated our FPCB picocalorimeter layout using Nanolithography Toolbox software, resulting in an output file in graphical design system (GDS) format. The layout design comprised two Cu and PI. A complete 38-line script for the device defined the soldering pads for the temperature sensor, the PI substrate with Cu lead-out, and the entire picocalorimeter (Supplementary Fig. 6). This script-based layout-generating approach enables direct modifications to the design by adjusting the script, simplifying the drawing process. Once we optimized the script, we created the GDS file and opened it in K-Layout software to verify the chip layout. Subsequently, we converted the GDS file into Gerber plot data format and sent it to the PCB manufacturer (see Supplementary Note 7). ### Measurement platform We constructed a system featuring three layers of thermal shields with significant thermal mass to decouple high-frequency temperature fluctuations from the ambient environment, enhancing the picocalorimeter temperature resolution. We regulated the temperatures of each shield using temperature control systems equipped with a PID feedback loop to minimize low-frequency temperature fluctuations. (Supplementary Figs. 3a, b) The chamber consists of two distinct temperature-controlled structures, with a total vacuum enclosure volume of ≈ (22 × 22 × 9.6) cm^3^ and a wall thickness of ≈ 2.0 cm, featuring electrical, optical, and feed-throughs that maintain the vacuum below 10 mPa. The outer chamber includes three connectors (KF25) for vacuum and electrical connections, along with two glass windows for optical access. Thermistors are bonded to the outer shield using a Cu holder at a representative location with epoxy, providing temperature feedback for PID control. Two thermoelectric coolers, each measuring ≈ (4 × 4) cm^2^, are attached in series to the outer chamber’s rear surfaces to regulate heating and cooling within the system. The middle chamber has dimensions of ≈ (14.5 × 14.5 × 4.2) cm^3^ and a wall thickness of about 1.0 cm, featuring two glass windows for optical inspection of the sample. The inner chamber measures ≈ (12.3 × 12.3 × 2.9) cm^3^ and has a wall thickness of ≈ 0.5 cm, incorporating two ≈ 1.0 cm-diameter openings fashioned from black ABS, which has thermal conductivity ranging from ≈ 0.14 W·m^−1^·K^−1^ to ≈ 0.21 W·m^−1^·K^−1^, and serves to isolate thermal fluctuations. The inner chamber is mounted directly within the middle chamber. Several modifications have been made to our upright optical microscope setup, enhancing our capacity to capture real-time video footage of the FPCB picocalorimeter. Among these modifications was the addition of a spacer that elevated the upper section of the microscope body. The temperature-stabilized chamber was placed on a heavy-duty, industrial-grade stage integrated into the microscope. This arrangement allowed us to focus on the picocalorimeter within the chamber using a 5× objective lens with a numerical aperture of 0.5, resulting in a total optical magnification of 50×. Either off-axis illumination (dark field) or differential interference contrast methods were often employed. A commercial camera with a resolution of 1920 × 1080 pixels was linked to the microscope via a c-mount adapter to capture the real-time video. Video recording and thermal measurements were conducted, followed by video analysis using *Adobe Premiere* software (see Supplementary Note 4). ### Device thermal parameters determination We quantify the *G* of our FPCB picocalorimeter system by applying a square signal, which induces a DC temperature rise (*T*~DC~) in the sensing area corresponding to the joule heat of *P* = (*V*~B~/2)^2^/*R*~0~, where *R*~0~ is the nominal electrical resistance of the thermistor at 25 °C, ≈ 100 kΩ. The same thermistor employs the aforementioned thermometry scheme with a bandwidth of 12.5 mHz to measure the resulting temperature rise. The slope of *P* with respect to *T* corresponds to the *G* value of the picocalorimeter system. We varied *V*~B~ from 0.8 V to 2 V in steps of 0.2 V, resulting in a *P* value ranging from ≈ 1.38 µW to ≈ 8.65 µW. We extracted corresponding *T*~DC~ values ranging from ≈ 14.52 mK to ≈ 81.30 mK. We plotted *T*~DC~ as a function of *P*. The d*P*/d*T* represents the picocalorimeter *G* value as (109.0 ± 8.2) µW·K^−1^ and (334.6 ± 32.9) µW·K^−1^, both (mean ± fitting error) in vacuum and air, respectively. Subsequently, we applied a light pulse from an LED while recording *V*~L~ as a function of *t* (Fig. 3c). We performed nonlinear curve fitting using a first-order exponential function to obtain the value of τ as (4.057 ± 0.001) s and (1.106 ± 0.001) s, both (mean ± SD) in vacuum and air, respectively (see Supplementary Note 3). ### Data analysis We implemented the transfer function relating measured voltage to heat power within a model based on the heat balance equation in MATLAB’s environment. The Δ*V* to Δ*T* is determined by the transfer equation of the Wheatstone bridge defined 4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \Delta T=\frac{4}{{{\rm{\alpha }}}\cdot {V}_{{{\rm{B}}}}}\Delta V, $$\end{document}ΔT=4α⋅VBΔV,where the transfer equation remained a constant number when a known *V*~B~ was applied. The calculation of *P* from *T* was a transient process, represented by a heat balance equation that evaluated the system *C*~p~ using the *τ* value, calculated as5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ P={C}_{{{\rm{p}}}}\frac{{{\rm{d}}}\Delta T}{{{\rm{d}}}t}+G\cdot \Delta T, $$\end{document}P=CpdΔTdt+G⋅ΔT, Our system exhibited a *τ* value of (4.057 ± 0.001) s and (1.106 ± 0.001) s, representing the mean ± SD under vacuum and ambient pressure, respectively, as detailed in the “Methods” section. We established those parameters and recalculated the input *P* generated from the sample in MATLAB (Fig. 4a) (for equations, refer to Supplementary Note 1 and 3). ### BSA solution preparation We prepared a ≈ 100 mg·mL^−1^ BSA solution in DI water for the following experiment. We added 1 mL of DI water into a sterilization centrifuge tube. Then, we weighed out 100 mg of BSA powder using a calibrated analytical balance and placed it into the tube with DI water. Gently swirl or stir the mixture to ensure that the BSA dissolves completely, followed by a filtration process. We also prepared the stock KCl solution with a concentration of ≈ 4.5 M by dissolving a specific amount of KCl in deionized water at ≈ 50 °C. The solution was then cooled to ≈ 25 °C to achieve saturation. This concentrated solution served as the starting point for further dilution, allowing us to conduct experiments at various concentrations ranging from ≈ 4 M to ≈ 15.6 mM. We then prepared four samples, including DI water and KCl solutions, with concentrations of ≈ 0.0625 M, ≈ 0.25 M, and ≈ 1 M KCl, each containing ≈ 100 g·L^−1^ of BSA in a 1 ratio. ## Supplementary information Supplementary Information Transparent Peer Review file ## Source data Source Data