使用正交實驗法探討化學共沉製程參數對Fe3O4奈米粒子交流磁場加熱的影響

Abstract

本研究使用化學共沉法合成Fe3O4奈米粒子, 在約11.3 Oe, 87 kHz的交流磁場中測量升溫曲線, 並使用牛頓冷卻模型擬合出樣品的比吸收率, 進而比較了不同製程參數組合的樣品的比吸收率之間的差異.

具體而言, 本研究涉及的化學共沉法的製程參數有: (1) Fe3+/Fe2+ 莫耳比 [1.2, 1.4, 1.6, 1.8]; (2) 通過滴定反應時長 [69, 46, 35, 28 (min)] 控制生成Fe3O4奈米粒子的速率 [0.4, 0.6, 0.8, 1.0 (μmol/s)]; (3) 溶液的體積 [40, 50, 60, 70 (mL)]. 在這個範圍內, 若採用全面實驗, 則有64個實驗組, 若每個實驗組還要3~5重覆 (或以上) 使數據穩定, 則實驗量非常龐大.

為了通過少量的實驗找出該範圍內的比吸收率變化規律, 本研究設計了正交實驗法, 從64個實驗組中抽取典型的16個進行重覆實驗直至平均值趨於穩定 (相對誤差不超過5 %). 對數據進行極差分析, 得出了前述三個參數各自對比吸收率的影響, 篩選並通過實驗確認了該範圍內的最優解為 [1.2, 0.8 μmol/s, 60 mL], 此時的最大比吸收率實驗值為31.60 W/g.

接着, 以16個正交實驗組為基準, 結合極差分析的意義, 建立了計算模型, 預測全面實驗的數據, 並初步評估了該方法的預測誤差約為11.27 %, 而在最優解附近的平均預測誤差約為5.76 %, 減少了大約一半. 該方法得出的預測值與正交實驗組之間的關係為: 兩者的極差分析表幾乎相同. 本研究還評估了數據迭代對預測誤差的影響: 迭代雖然會使預測數據穩定, 但也累積了誤差.

未來, 本研究找出的最優解可進行表面功能化, 應用於具體的生醫領域 (例如, 癌症熱療, 靶向藥物輸送等).

In this study, Fe3O4 nanoparticles were synthesized using chemical co-precipitation method. The temperature rise curves were measured in an AC magnetic field of approximately 11.3 Oe and 87 kHz. The specific absorption rate (SAR) of the samples was fitted using the Newton’s cooling model, allowing for a comparison of SAR among samples with different process parameter combinations.

Specifically, the process parameters involved in the chemical co-precipitation method were: (1) Fe3+/Fe2+ molar ratio [1.2, 1.4, 1.6, 1.8]; (2) rate of generating Fe3O4 nanoparticles [0.4, 0.6, 0.8, 1.0 (μmol/s)] controlled by titration reaction duration [69, 46, 35, 28 (min)] ; and (3) solution volume [40, 50, 60, 70 (mL)]. Within this range, conducting a comprehensive experiment would entail 64 experimental groups, with each group requiring 3 to 5 repetitions (or more) for data stability, resulting in a substantial number of experiments.

In order to identify the variation pattern of SAR within this range through a small number of experiments, this study employed orthogonal experimental design. 16 typical experimental groups were selected from the 64, and repeated experiments were conducted until the average value stabilized (with a relative error not exceeding 5%). Range analysis of the data determined the influence of each of the three parameters on SAR. The optimal solution within this range was identified as [1.2, 0.8 μmol/s, 60 mL], yielding a maximum SAR of 31.60 W/g.

Subsequently, using the 16 orthogonal experimental groups as a reference and considering the significance of range analysis, a computational model was established to predict the data from comprehensive experiments. The method's prediction error was initially estimated to be approximately 11.27%, which was reduced by about half near the optimal solution, resulting in an average prediction error of about 5.76%. The relationship between the predicted values obtained by this method and the orthogonal experimental groups closely matched the range analysis table. This study also assessed the impact of data iteration on prediction error, noting that while iteration stabilized the predicted data, it also accumulated error.

In the future, the optimal solution identified in this study can be subjected to surface functionalization for specific biomedical applications, such as cancer hyperthermia and targeted drug delivery.