Research Progress of Silicon-based BIB Infrared Detector
-
摘要: 以锗基和硅基为主的阻挡杂质带(blocked impurity band,BIB)红外探测器的兴起有力推进了红外天文学的快速发展,其中硅基BIB红外探测器在特定波长的航天航空领域有着不可替代的地位。国外对硅基BIB红外探测器的研究已有40多年,以美国航空航天局(NASA)为主的科研机构已经实现了硅基BIB红外探测器在天文领域的诸多应用,而国内对硅基BIB红外探测器的研究尚处于起步阶段。本文首先阐述了硅基BIB红外探测器的工作原理,然后简单概述了器件结构和制备工艺,并对不同类型的硅基BIB探测器的性能进行了对比分析,之后介绍了其在天文探测中的应用,最后对硅基BIB红外探测器未来的发展进行了展望。Abstract: The rise of blocked impurity band (BIB) infrared detectors based on germanium and silicon has promoted the rapid development of infrared astronomy, among which silicon-based BIB infrared detectors with specific wavelengths play an irreplaceable role in the aerospace field. Research on silicon-based BIB infrared detectors has been conducted abroad for more than 40 years, and many of its applications in the astronomical field have been realized by NASA and its related research institutes. However, domestic research on silicon-based BIB infrared detectors is still in its infancy. In this paper, the working principle of silicon BIB infrared detectors is described first; then, the structure and fabrication process of the device are briefly summarized, the performance of different types of silicon BIB detectors is compared and analyzed, and its application in astronomical detection is described. Finally, the future development of silicon BIB infrared detectors is discussed.
-
Keywords:
- silicon-based BIB /
- infrared detector /
- astronomical detection
-
0. 引言
近年来,借助于深度学习等理论,视觉目标跟踪技术取得了重大突破。但现有跟踪算法大多是针对可见光场景,对于热红外目标跟踪的研究文献较少,只有LIU等人[1-4]进行了较为深入的研究。
与可见光目标跟踪相比,热红外跟踪具有不受照明变化影响的优点,可以在完全黑暗的情况下跟踪目标。因此,热红外跟踪在灾难救援、视频监控和夜间巡逻等领域具有广阔的应用前景[5]。随着热红外成像设备在民用领域的普及化使用,热红外目标跟踪正逐步成为计算机视觉领域的一个研究热点。
和可见光跟踪相似,热红外跟踪也面临着不少挑战。首先,热红外图像缺少颜色信息,较难获得目标对象的判别特征,从而降低了跟踪性能[6]。比如两个不同颜色的相似物体在热红外图像中几乎是相同的。除此之外,热红外跟踪还面临如遮挡、外观变化和运动模糊等挑战。
近年来,深度学习已成功应用到视觉跟踪中,取得了一系列重要成果。与基于手工特征的跟踪器相比,基于卷积神经网络(Convolutional Neural Networks,CNN)的跟踪器可以获得更优越的跟踪性能。因为单个卷积层特征对跟踪中的各种挑战(例如形变和背景杂斑)鲁棒性不强,现有基于深度学习的跟踪算法大多是融合浅层和深层特征进行跟踪,比如HCF(Hierarchical Convolutional Features)[7]基于KCF(Kernel Correlation Filter)框架,使用预训练的VGG-19(Visual Geometry Group-19)网络提取目标的Conv3-4(Convolution 3-4)、Conv4-4、Conv5-4层特征进行融合,而TADT(Target-Aware Deep Tracking)[8]基于全卷积孪生网络(Fully-Convolutional Siamese Network,SiamFC)[9]框架使用预训练的VGG-16网络提取Conv4-1、Conv4-3层特征并进行融合等等。上述算法均属于特征层融合,并没有实现决策层融合,而且使用单一跟踪器无法有效应对复杂背景及目标外观变化等挑战因素。
考虑到卷积神经网络的强大表征能力及其在视觉跟踪中的成功应用,本文将在可见光图像数据集上预训练的CNN用于热红外跟踪任务,并基于此设计了一个多响应图集成的热红外目标跟踪方法。
1. 全卷积孪生网络框架
全卷积孪生网络是ECCV 2016(European Conference on Computer Vision 2016)会议[9]上提出的一种全新的目标跟踪框架,其通过比较初始目标模板和当前帧中搜索区域的特征来进行跟踪。其实现流程可以由下式定义:
$$ f\left( {z, x} \right) = \varphi \left( z \right)*\varphi \left( x \right) + b $$ (1) 式中:ϕ(z)为第一帧中模板图像z对应的特征图;ϕ(x)为后续帧中待搜索区域x的特征图;*为卷积运算符,b为偏置项。图 1给出了SiamFC网络结构图。
本文算法选择全卷积孪生网络作为目标跟踪的基本框架,主要原因在于:①全卷积孪生网络结构简单,可以在较大的搜索图像中定位目标模版,并且跟踪速度快。②在跟踪过程中目标模板不更新,可以有效应对目标遮挡等挑战。
2. 本文算法
2.1 预训练网络
虽然Siamese框架是在大型视频数据集ILSVRC15(Large Scale Visual Recognition Challenge 2015)上预先训练的,但是Siamese跟踪器没有充分利用与特定目标对象相关的语义信息。因此在本文算法中,我们借鉴了TADT的做法,从一个预先训练好的CNN中选择对于目标最具判别性以及对于目标尺度变化最为敏感的卷积滤波器(特征通道)来生成目标深度特征。在本文中,我们选择使用VGG-16卷积网络来提取对于热红外目标敏感的特征。在2.2节中将给出特征通道的选择方法。
2.2 通道选择
诸如VGG、ResNet(Residual Network)等卷积网络中的参数是在庞大的数据集中预训练获得的,相对于当前要跟踪的对象存在许多冗余的特征通道信息。冗余的特征通道会占用大量的计算资源,这也是许多基于深度学习的跟踪算法执行速度较慢的原因。除此之外,冗余的特征通道还会对跟踪造成干扰,影响跟踪精度。本文在特征提取过程中进行通道选择,减少冗余特征通道对当前跟踪目标的不利影响。
本文的通道选择是基于文献[10]提出的GCAM(Grad-Class Activation Mapping)模型来实现的。GCAM模型根据每个输入像素点的梯度来表示其属于给定类别标记的重要性,通过沿特征通道计算加权和,生成一个类活动映射。特征通道的权值是通过对该通道中所有梯度的全局平均池化(Global Average Pooling,GAP)来计算的,权值高的特征通道对当前类的贡献大。公式如下:
$$ {\varDelta _i} = {G_{{\text{AP}}}}\frac{{\partial L}}{{\partial {z_i}}} $$ (2) 式中:Δi表示计算出的第i个通道的重要性;GAP是全局平均池化函数;L是损失函数;zi指的是第i个过滤器的输出特征。
借鉴该理论,本文算法在提取深度特征后计算每个特征通道的梯度值,从中选择最能标记目标特征也就是重要性较高的特征通道进行后续运算。
2.3 多跟踪器设计
虽然当前主流跟踪算法都使用了深度特征,但使用单个跟踪器在某些场景下仍然很难获得优秀的跟踪性能。基于此,本文提出利用多个互补的卷积层特征构建一个集成式热红外目标跟踪方法。本文利用VGG-16卷积网络分别提取待跟踪目标的Conv4-1、Conv4-2、Conv4-3和Conv5-1层特征,并形成如下3种特征组合:{Conv4-2、Conv4-1}、{Conv4-3、Conv4-1}、{Conv5-1、Conv4-1}。利用上述3种特征组合分别设计跟踪器1、跟踪器2、跟踪器3,在3种特征组合中均包含Conv4-1层,这是因为文献[8]的实验表明Conv4-1层特征最适合用来进行目标的尺度估计。每个跟踪器的计算过程都是相似的,下面以跟踪器1为例给出具体跟踪过程:
1)提取待跟踪目标的Conv4-1和Conv4-2层特征(每层共包含512个特征通道);
2)利用2.2节中提到的方法计算所有特征通道的梯度信息,选择对于目标活动和尺度变化较为敏感的特征通道参与后序的互相关操作。经过通道选择后,降低了卷积层的维度,提高了卷积层的有效性。本文按照重要性高低对所有特征通道进行排序并选择了Conv4-2层的前300个通道和Conv4-1层的前80个通道进行融合参与后续运算。
3)通过互相关操作获取目标响应图。
2.4 基于KL散度的集成模型
利用2.3节的方法可以获得多个响应图,在本节将给出利用KL散度将多个响应图进行集成的过程。
我们可以用Pk表示第k个跟踪器产生的目标位置响应图,Q∈RM×N表示集成后的响应图。然后采用KL散度度量他们之间的广义距离,该距离越小则认为他们之间的分布差异越小,也就意味着集成后的概率图Q更能反映每一个概率图Pk中一致的部分,即目标出现概率最大的区域。通过最小化该KL散度,可以达到优化概率图Q的目的,具体过程如下式:
$$ \arg \mathop {\min }\limits_Q \sum\limits_{k = 1}^n {{\text{KL}}({P^k}||Q)} $$ (3) $$ {\text{s}}{\text{.t}}{\text{.}}\quad \sum {{q_{ij}} = 1} $$ 其中:
$$ {\text{KL}}({P^k}||Q) = \sum\limits_{ij} {p_{ij}^k} \log \frac{{p_{ij}^k}}{{{q_{ij}}}} $$ (4) 式中:pij,qij分别表示概率图P与概率图Q中的第(i, j)个元素。pijk表示概率图Pk中目标位置为(i, j)的概率,i={1, 2, …, M}, j={1, 2, …, N},并且服从$ \sum {p_{ij}^k} = 1 $。
由于在跟踪场景中可能存在背景杂斑等干扰,因此在响应图中会存在“多峰”现象,即存在多个目标位置响应。文献[1]提供了一个行之有效的去噪策略,本文借鉴该策略将2.3节中得到的响应图进行过滤,下式给出过滤过程:
$$ {P^{k, z}} = {P^k} \odot {P^z} $$ (5) 式中:k={1, 2, …, n-1},z={k+1, k+2, …, n}。
在式(5)中n代表跟踪器的数量;⊙表示点乘操作。按照式(5)进行去噪处理可以使过滤后的概率图具有更高的置信度,提高跟踪的精确度。n个响应图进行两两去噪处理后可以得到(n(n-1))/2个噪音更少、置信度更高的概率图A={P1, 2, P1, 3, …, P2, 3, …Pn-1, n}。因此公式(3)可以被重写成如下形式:
$$ \arg \mathop {\min }\limits_Q \sum\limits_{p \in A} {\sum\limits_{ij} {{p_{ij}}} } \log \frac{{{p_{ij}}}}{{{q_{ij}}}} $$ (6) $$ {\text{s}}{\text{.t}}{\text{.}}\quad \sum {{q_{ij}} = 1} $$ 根据拉格朗日乘数法,令:
$$ F({p_{ij}}, {q_{ij}}) = \sum\limits_{p \in A} {\sum\limits_{ij} {{p_{ij}}} } \log \frac{{{p_{ij}}}}{{{q_{ij}}}} + \lambda (\sum {{q_{ij}} - 1} ) $$ (7) 对所有的pij、qij求偏导并令F′(p)和F′(q)等于0,最终可求得:
$$ Q = \frac{2}{{n(n - 1)}}\sum\limits_{p \in A} p $$ (8) 通过取式(8)中响应图Q中的最大值可以得到目标的位置(x, y):
$$ (x, y) = \arg \mathop {\max }\limits_{i, j} Q(i, j) $$ (9) 3. 实验结果与分析
3.1 实验环境
我们在Win10系统中使用Matlab2016b编程语言实现了所提出的目标跟踪算法。为了验证本文所提算法的有效性,在配备I7-10700 2.9 GHz CPU和GTX-1080 GPU的PC上进行了实验,对比算法有HCF[7]、SiamFC[9]、TADT[8]、MDNet(Multi-Domain Convolutional Neural Networks)[11]、ATOM(Accurate Tracking by Overlap Maximization)[12]、MCFTS(Multi-Layer Convolutional Features for Thermal Infrared Tracking)[1]、HSSNet(Hierarchical Spatial-Aware Siamese Network)[2]、MLSSNet(Multi-Level Similarity Network)[3]、MMNet(Multi-Task Matching Network)[4]。其中MCFTS、MLSSNet、HSSNet和MMNet是专用红外目标跟踪算法,其他5种算法均为近年来优秀的可见光跟踪算法。
3.2 评估数据集
因为目前公开的红外视频数据集非常少,所以很多学者使用VOT-TIR16(The Thermal Infrared Visual Object Tracking 2016)[13]和PTB-TIR(Thermal Infrared Pedestrian Tracking Benchmark)[14]作为热红外跟踪评估数据集。VOT-TIR16数据集中视频序列过少,只包含25个红外视频序列,部分跟踪算法可以通过调参的方式去适应。PTB-TIR数据集中虽然包含60个红外视频序列,但跟踪目标均为行人,类别较为单一,无法评估跟踪算法对通用目标的跟踪效果。
2020年8月发布的LSOTB-TIR[15]有效解决了上述问题。LSOTB-TIR是迄今为止规模最大,多样性最强的热红外目标跟踪数据集。包含120个跟踪序列,涵盖22个目标类别,超过8.2万帧图像。LSOTB-TIR中定义了4种场景属性,这4种场景中的红外视频分别由车载、手持、无人机载和监控设备拍摄,保证了视频来源的多样性。著名的可见光数据集OTB-100(Object Tracking Benchmark)定义了11种挑战属性,LSOTB-TIR则定义了红外目标跟踪中的12种挑战属性。在这12种挑战属性中,有8种属性和OTB-100中的类似,分别为遮挡、出视野、背景杂斑、形变、尺度变化、低分辨率、快速运动和运动模糊,这里不再赘述。剩余的4种为热红外跟踪中特有的挑战属性,具体定义在表 1中列出。
表 1 LSOTB-TIR定义的的4种热红外挑战属性Table 1. Four thermal infrared challenge attributes defined by LSOTB-TIRInfrared challenge attributes Specific definitions Aspect Ratio
Variation(ARV)The aspect ratio of the target exceeds [0.5, 2] during tracking Intensity Variation
(Ⅳ)The intensity of the target changes during tracking Thermal Crossover
(TC)Two targets of the same intensity cross each other Distractor(DIS) There are interfering objects similar to the target around the target 因为LSOTB-TIR的全面性和多样性,所以本文选择使用LSOTB-TIR作为评估数据集。需要说明的是,本文并没有选择全部视频序列进行测试,而是从上述4种应用场景数据集中随机选择了40个典型红外短视频序列进行测试,这是因为LSOTB-TIR没有跟踪失败重启机制,在长视频序列中如果在某一帧丢失目标并且在后续帧无法找回将极大影响跟踪成功率,采用短视频序列测试更加公平。本文所提算法在40个视频序列上的平均跟踪速度约为11.2 FPS。
3.3 与经典跟踪算法的性能对比
3.3.1 定量分析
图 2为本文算法与MDNet等9种对比算法在LSOTB-TIR上生成的总精确度图和成功率图,本文提出的算法在这两项重要指标上均排名第一。与参照算法TADT相比,跟踪精确度和成功率分别提高了3.6%和4.2%,大幅超过了MMNet等红外目标跟踪算法,证明了本文采用的响应图集成策略取得较好的效果。图 3给出了10种算法在变形、背景杂斑、运动模糊、相似物干扰等4种复杂场景下的精确度曲线图,本文所提算法均体现了较强的鲁棒性。相比优秀的可见光跟踪器ATOM,本文算法的跟踪精确度率分别提高了13.2%、10.5%、13.9%和27.8%,说明本文所提算法更适合于红外目标跟踪。
3.3.2 定性分析
为了更直观地评估本文所提算法的跟踪性能,图 4给出了本文算法与另外3种在LSOTB-TIR数据集上表现优异的算法(TADT、ATOM和MDNet)在car_S_003、cow_H_001、person_S_015、street_S_001等4个视频序列的可视化跟踪结果。这4个视频序列非常具有代表性,包括汽车、动物、行人和环境复杂的街道,几乎涵盖了重要种类的地面红外目标。定性分析过程如下:
1)car_S_003视频序列中跟踪目标为行进中的汽车。ATOM跟踪器在第9帧时即受到周围相似物的干扰而发生漂移,第52帧中TADT和MDNet也接近丢失目标,第105帧时只有本文算法可以准确跟踪。
2)cow_H_001跟踪目标为右起第二只羊,从视频截图可以看出,ATOM和MDNet均受到相似物和周围环境干扰丢失目标,只有本文算法和TADT可以持续准确跟踪。
3)person_S_015视频序列中跟踪目标为右边的行人,旁边有个撑伞的行人和其同步前进,存在相似物及热交叉等干扰。ATOM、TADT分别在第30帧、第210帧时跟踪失败。相比MDNet,本文算法定位更为精确。
4)street_S_001视频序列中跟踪目标为街道上背包的行人,该视频序列包含热交叉、背景杂斑、遮挡等多种挑战。从视频截图可以看出其他几种算法跟踪框均漂移到相似的行人上,而本文算法因为集成了多个跟踪器可以正确跟踪。
3.4 消融实验
本节通过标准数据集LSOTB-TIR上的消融实验来验证算法相关模块设计的合理性。首先,为了验证通道选择的有效性,以跟踪器1(Ours_1)为例,比较了不进行通道选择(No_CS)和进行通道选择的效果对比。如图 5(a)所示,本文算法在只选择了300个通道的情况下跟踪精度还提高了0.6%,证明冗余通道可以丢弃,并且丢弃冗余通道后跟踪速度有了显著提升,因此利用通道选择后的跟踪器参与后续集成更加高效。其次,为了验证集成方法的有效性,将本文提出的集成方法(Ours)与跟踪器1(Ours_1)、跟踪器2(Ours_2)及跟踪器3(Ours_3)分别单独跟踪的结果进行了对比。如图 5(b)所示,可以看出集成算法较其他3个单独跟踪方法在跟踪精确度上分别提升了2.5%、3.6%和1.6%。消融结果证明了本文所提算法的有效性。
4. 结论
基于简洁高效的全卷积孪生网络框架提出了一种多响应图集成的红外目标跟踪算法。算法的创新之处在于利用预训练的VGG-16网络来提取热红外目标的深度特征并进行通道选择,进而通过对不同层次的深度特征进行组合设计了3个跟踪器。最后基于KL散度将3个跟踪器的响应图进行集成以获取最优结果。为了验证本文算法的实际效果,在大型热红外数据集LSOTB-TIR上进行了测试并和近年来几种先进的目标跟踪算法进行了比较。实验结果表明,本文所提算法性能良好,具有一定的实际应用价值。
-
图 2 硅基BIB红外探测器的结构和工作原理:(a) 非本征硅光电导探测器的工作原理示意图[10];(b) 硅基BIB红外探测器的工作原理图[11];(c) Si: As BIB红外探测器结构示意图[13];(d) Si: Sb BIB红外探测器的器件结构图[14];(e) 背照射式Si: Sb BIB探测器的结构示意图,其中Nd为中性施主的密度,Nd+为电离施主的浓度,Na-为电离受主的浓度[15];(f) Si: Sb BIB探测器的红外吸收层在正的反偏电压下的平衡电荷分布图[15]
Figure 2. Structures and working mechanisms of silicon-based BIB infrared detectors: (a) Schematic diagram of the working principle of the ESPC detector[10]; (b) Schematic diagram of the working principle of the silicon-based BIB infrared detector[11]; (c) Structure diagram of the Si: As BIB infrared detector[13]; (d) Structure diagram of the Si: Sb BIB infrared detector[14]; (e) Schematic diagram of the back-illuminated Si: Sb BIB, where Nd is the density of neutral donors, Nd+ is the ionized donor density, and Na- is the density of ionized acceptors[15]; (f) Equilibrium charge distributions for the positive reverse-biased operation for the Si: Sb BIB infrared detector[15]
图 3 硅基BIB红外探测器的性能:(a) 用于Si: As IBC探测器辐射测试的低温杜瓦装置[38];(b) 测试及计算得到的Si: As IBC探测器的响应量子效率曲线[38];(c) Si: As IBC探测器的I-V测试曲线[38];(d) 金属管壳封装的Si: Sb BIB探测器[14];(e) Si: Sb BIB探测器的光谱量子效率曲线[14];(f) Si: Sb BIB探测器的暗电流与温度的关系[14];(g) Si: P BIB器件的PC光谱与远红外背景光谱,以及响应峰的指定[39];(h) Si: Ga BIB探测器的光谱量子效率[40];(i) Si: Ga BIB探测器与长波碲镉汞探测器的暗电流对比[40]
Figure 3. Performances of the silicon-based BIB infrared detectors: (a) Dewar configuration for Si: As IBC detector radiation testing[38]; (b) Responsive quantum efficiency curves of Si: As IBC detector[38]; (c) I-V testing curves of Si: As IBC detector[38]; (d) Metal shell packed Si: Sb BIB detector[14]; (e) Spectral quantum efficiency curve of Si: Sb BIB detector[14]; (f) Dark current as a function of temperature of Si: Sb BIB detector, measured at 1.5 V bias voltage[14]; (g) PC spectrum of the Si: P BIB device versus far-infrared background spectrum, and the designations of the response peak[39]; (h) Spectral QE of Si: Ga BIB detector[40]; (i) Dark current performance comparison of Si: Ga BIB detector with LWMCT detector[40]
图 5 国外硅基BIB红外探测器的研究进展:(a) 空间红外望远镜设备(SIRTF)上的128×128长波长红外焦平面组件[29];(b) DRS公司的HF1024焦平面阵列,封装在84针无铅芯片载体上[40];(c) 百万像素中红外阵列裸多路复用器[54];(d) 无掺杂单晶衬底晶圆[54];(e) Si: As BIB焦平面阵列的封装[55];(f) 256×256 Si: As IBC阵列及其航天封装[57];(g) 1024×1024 Si: As IBC阵列的红外传感器芯片[53];(h) 1024×1024 Si: As IBC阵列的读出电路[58];(i) 由双侧可粘扣的HF1024 Si: As和Si: Sb焦平面阵列组成的2048×2048焦平面阵列,像元间距为18 μm[40]
Figure 5. Research progresses of overseas silicon-based BIB infrared detectors: (a) SIRTF 128×128 long wavelength infrared focal plane array assembly[29]; (b) DRS HF1024 FPA packaged in 84-pin leadless chip carrier[40]; (c) A Mega pixel MIR bare multiplexer[54]; (d) Undoped single-crystal substrate wafer[54]; (e) Packaging of the BIB focal plane arrays[55]; (f) 256×256 Si: As IBC array in flight mount[57]; (g) Photo of a 1024×1024 Si: As IBC SCA[53]; (h) SB-291 ROIC for 1024×1024 Si: As IBC array[58]; (i) 2048×2048 FPA with 18-micron pixel pitch composed of 2-side buttable HF1024 Si: As and Si: Sb FPAs[40]
图 6 国内硅基BIB红外探测器的研究进展:(a) 平面型Si: P BIB探测器结构示意图[65];(b) 垂直型Si: P BIB探测器模型[58];(c) Si: P BIB探测器在2 V偏压和不同温度下的响应光谱[58];(d) 等离子体调谐太赫兹探测器横截面示意图[59];(e) 不同周期性孔结构(PHSs)的Si: P BIB探测器的归一化光电流谱[59];(f) Si: Ga BIB探测器在不同功能区上的层状材料结构示意图[60];(g) Si: Ga BIB探测器不同温度下的响应谱[60];(h) 金属光栅/硅基BIB太赫兹探测器的工作原理图[61];(i) 有金属光栅的器件(参数:p=7 μm,d=5 μm,DR=2/7)与无金属光栅的器件的实验光谱响应对比[61]
Figure 6. Research progresses of domestic silicon-based BIB infrared detectors: (a) Schematic diagram of the planar type Si: P BIB detector structure[65]; (b) Vertical type Si: P BIB detector model[58]; (c) Response spectrum of the Si: P BIB detector at 2 V bias voltage with different temperatures[58]; (d) Schematic representation of the cross section of the plasma-tuning THz detector[59]; (e) The normalized photocurrent spectrum of the Si: P BIB detectors for different periodic pore structures (PHSs)[59]; (f) Schematic diagram of the layered material structure of the Si: Ga BIB detector in different functional areas[60]; (g) Response spectrum of the Si: Ga BIB detector at different temperatures[60]; (h) Mechanism of the metal-grating/silicon-based BIB THz detector[61]; (i) Comparison of the experimental spectral response of devices with metal gratings (parameters: p=7 μm, d=5 μm, DR=2/7) with devices with metal-free gratings[61]
图 7 硅基BIB红外探测器的天文应用[72]:(a) 斯皮策太空望远镜;(b) 斯皮策太空望远镜观测到的“红蝴蝶”星系;(c) WISE捕捉的最古老的超新星RCW 86的图像;(d) 水瓶座/SAC-D航天探测器;(e) 平流层天文台;(f)平流层天文台捕捉的恒星合并的快照;(g) 詹姆斯·韦伯空间望远镜(JWST);(h) JWST的近红外照相机捕捉的第一张全彩图像;(i) COBE在太空中运行的示意图
Figure 7. Astronomical applications of the silicon-based BIB infrared detectors[72]: The spitzer space telescope; (b) The "red butterfly" galaxy was observed by the spitzer space telescope; (c) An image of the oldest supernova RCW 86 captured by WISE; (d) The aquarius/SAC-D space probe; (e) Stratospheric observatory for infrared astronomy; (f) Snapshot of stellar mergers captured by SOFIA; (g) The James Webb Space Telescope; (h)The first full color image captured by the near-infrared camera of the JWST; (i) Schematic representation of the cosmic background explorer operating in space
表 1 硅基BIB红外探测器的部分工艺参数
Table 1 Partial process parameters of the silicon-based BIB infrared detector
Year Material Thickness of IRAL/μm Thickness of blocking layer/μm Doping concentration of IRAL/cm-3 Fabrication method of epitaxial layer Institution Ref. 1979 Si: As 6−10 1−4 7×1017 CVD Rockwell [11] 1992 Si: Sb 17 3.5 1−8×1017 CVD Rockwell [15] 1999 Si: B 4.5 3 1×1018 - - [16] 2007 Si: As 10 - 4×1018 - DRS [17] 2007 Si: P - - 4×1018 - DRS [17] 2018 Si: As 15 - 1×1018 - NIST [18] 表 2 国外公司生产的硅基BIB红外探测器的性能参数
Table 2 Performance parameters of silicon-based BIB infrared detectors produced by foreign companies
Year Material Technology FPA format Pixel size/μm2 Pixel pitch/μm Operating temperature range/K Wavelength
range/μmDark current Quantum efficiency/% Institution Applications Ref. 2012 Si: Sb BIB 1024×1024 18 - 5-12 14-38 0.1 e/s 60 DRS Wide-field infrared survey explorer [14] 1992 Si: Sb BIB 128×128 - - 7 2-40 - - Rockwell Space infrared telescope facility [15] 2018 Si: As BIB - - - 7-10 2-30 10-12 A/mm2 60 NIST Missile defense transfer radiometer [18] 1986 Si: As BIB 10×50 - - 12 - 12.3 pA - Rockwell - [19] 1991 Si: As BIB 128×128 75 - 11 - < 0.1 nA - Rockwell Space infrared telescope facility [20] 1993 Si: As BIB 256×256 30 - 12 - 18 e-/s 57 HTC Space infrared telescope facility [21] 1995 Si: Ga ESPC 128×192 75 - ≤10 5-17 0 30 LETI/LIR European transonic windtunnel [22] 1998 Si: As BIB 256×256 30 - 6-7 - < 100 e-/s 40 RVS Infrared imaging surveyor [23] 1998 Si: As BIB 320×240 - 50 6 2-28 100 e-/s 40-55 SBRC SUBARU [24] 2000 Si: As BIB 256×256 - 25 6 5-28 3.8 e-/s 84 RVS Space infrared telescope facility [25] 2001 Si: As BIB 320×240 - 50 6-12 2-28 ≤100 e-/s > 40 RVS Mid-infrared spectrometer and imager [26] 2001 Si: As BIB 1024×1024 - 27 6-8 5-30 0.3 e-/s 45 RVS Next generation space telescope [27] 2001 Si: As BIB 1024×1024 - 27 6 5-30 < 1 e-/s 50 RVS Stratospheric observatory for infrared astronomy [28] 2003 Si: As BIB 128×128 - 75 - - 0.49-2.9 e-/s 84 DRS Wide-field infrared explorer [29] 2003 Si: Sb BIB 128×128 - 75 - - 5.3-12.9 e-/s 51 DRS Wide-field infrared explorer [29] 2003 Si: As BIB 256×256 50 - 4.7 5-25 - 56 DRS Stratospheric observatory for infrared astronomy [30] 2004 Si: As BIB 256×256 25 - 6.7-7.1 5-28 0.1 e-/s > 50 RVS James Webb space telescope [31] 2005 Si: As BIB 1024×1024 - 18 6 5-28 < 10 e-/s > 57 DRS Wide-field infrared survey explorer/James Webb space telescope [32] 2006 Si: As BIB 1024×1024 - 18 7.8 7.5-28 < 100 e-/s > 60 DRS Wide-field infrared survey explorer [33] 2008 Si: As BIB 1024×1024 30 - 7-9 3-28 1 e-/s > 40 RVS AQUARIUS [34] -
[1] McCreight C R, McKelvey M E, Goebel J H, et al. Detector arrays for low-background space infrared astronomy[C]//Infrared detectors, Sensors, and Focal Plane Arrays of SPIE, 1986, 686: 66-75.
[2] Szmulowicz F, Madarasz F L. Blocked impurity band detectors—an analytical model: figures of merit[J]. Journal of Applied Physics, 1987, 62(6): 2533-2540. DOI: 10.1063/1.339466
[3] Battersby C, Armus L, Bergin E, et al. The origins space telescope[J]. Nature Astronomy, 2018, 2(8): 596-599. DOI: 10.1038/s41550-018-0540-y
[4] 刘恩科, 朱秉升, 罗晋生. 半导体物理学: 7版[M]. 北京: 电子工业出版社, 2017. LIU Enke, ZHU Bingsheng, LUO Jingsheng. The Physics of Semiconductors: 7th Edition[M]. Beijing: Publishing House of Electronics Industry, 2017.
[5] CHEN H C, LIN C C, HAN H W, et al. Enhanced efficiency for c-Si solar cell with nanopillar array via quantum dots layers[J]. Optics Express, 2011, 19(105): A1141-A1147.
[6] JUANG J Y, ZHOU K, BANG J H, et al. Improved photovoltaic performance of Si nanowire solar cells integrated with ZnSe quantum dots[J]. The Journal of Physical Chemistry C, 2012, 116(23): 12409-12414. DOI: 10.1021/jp301683q
[7] WU C, Crouch C H, ZHAO L, et al. Near-unity below-band-gap absorption by microstructured silicon[J]. Applied Physics Letters, 2001, 78(13): 1850-1852. DOI: 10.1063/1.1358846
[8] Crouch C, Carey J, Shen M, et al. Infrared absorption by sulfur-doped silicon formed by femtosecond laser irradiation[J]. Appl Phys A, 2004, 79: 1635-1641. DOI: 10.1007/s00339-004-2676-0
[9] ZHANG T, Ahmad W, LIU B, et al. Broadband infrared response of sulfur hyperdoped silicon under femtosecond laser irradiation[J]. Materials Letters, 2017, 196: 16-19. DOI: 10.1016/j.matlet.2017.03.011
[10] 王占国, 郑有炓. 半导体材料研究进展[M]. 北京: 高等教育出版社, 2012. WANG Zhanguo, ZHENG Youliao. Research Progress in Semiconductor Materials[M]. Beijing: Publishing House of Higher Education, 2018.
[11] Petroff M D, Stapelbroek M G. Blocked Impurity Band Detectors: 4568 960[P]. U.S. Patent, 1986-02-04.
[12] Petroff M D, Stapelbroek M G. Responsivity and noise models of blocked impurity band detectors[C]//Proc. IRIS Specialty Group on Infrared Detectors, 1984, 2.
[13] Rieke G H. Infrared detector arrays for astronomy[J]. Annu. Rev. Astron. Astrophys. , 2007, 45: 77-115. DOI: 10.1146/annurev.astro.44.051905.092436
[14] Khalap V, Hogue H. Antimony-doped silicon blocked impurity band (BIB) arrays for low flux applications[C]//Infrared Sensors, Devices, and Applications Ⅱ. International Society for Optics and Photonics, 2012, 8512: 85120O.
[15] Huffman J E, Crouse A G, Halleck B L, et al. Si: Sb blocked impurity band detectors for infrared astronomy[J]. Journal of Applied Physics, 1992, 72(1): 273-275. DOI: 10.1063/1.352127
[16] Asadauskas L, Brazis R, Leotin J. Optical phonon line in boron-doped silicon BIB structures[C]//Materials Science Forum. Trans Tech Publications Ltd., 1999, 297: 361-364.
[17] Hogue H H, Guptill M T, Monson J C, et al. Far-infrared blocked impurity band detector development[C]//Infrared Spaceborne Remote Sensing and Instrumentation XV of SPIE, 2007, 6678: 63-73.
[18] Woods S I, Proctor J E, Jung T M, et al. Wideband infrared trap detector based upon doped silicon photocurrent devices[J]. Applied Optics, 2018, 57(18): D82-D89. DOI: 10.1364/AO.57.000D82
[19] Stetson S B, Reynolds D B, Stapelbroek M G, et al. Design and performance of blocked-impurity-band detector focal plane arrays[C]//Infrared Detectors, Sensors, and Focal Plane Arrays of SPIE, 1986, 686: 48-65.
[20] Noel R A. Large-area blocked-impurity-band focal plane array development[C]//Infrared Detectors and Focal Plane Arrays Ⅱ of SPIE, 1992, 1685: 250-259.
[21] Lum N A, Asbrock J F, White R, et al. Low-noise, low-temperature 256 ×256 Si: As IBC staring FPA[C]//Infrared Detectors and Instrumentation of SPIE, 1993, 1946: 100-109.
[22] Suffis S, Caes M, Deliot P, et al. Characterization of 128×192 Si: Ga focal plane arrays: study of nonuniformity, stability of its correction, and application for the CRYSTAL camera[C]//Infrared Detectors and Focal Plane Arrays Ⅴ of SPIE, 1998, 3379: 235-248.
[23] Matsuhara H. IRC: an infrared camera on board the IRIS[C]//Infrared Astronomical Instrumentation of SPIE, 1998, 3354: 915-921.
[24] Sohn E, Schneider E R, Cruz-Gonzales I, et al. Mid-infrared camera/spectrograph for OAN/SPM[C]//Infrared Astronomical Instrumentation, 1998, 3354: 822-824.
[25] McMurray Jr R E, Johnson R R, McCreight C R, et al. Si: As IBC array performance for SIRTF/IRAC[C]//Infrared Spaceborne Remote Sensing Ⅷ of SPIE, 2000, 4131: 62-69.
[26] Deutsch L K, Hora J L, Adams J D, et al. MIRSI: a mid-infrared spectrometer and imager[C]//Instrument Design and Performance for Optical/Infrared Ground-based Telescopes of SPIE, 2003, 4841: 106-116.
[27] Ennico K A, McKelvey M E, McCreight C R, et al. Large format Si: As IBC array performance for NGST and future IR space telescope applications[C]//IR Space Telescopes and Instruments of SPIE, 2003, 4850: 890-901.
[28] Ennico K A, Greene T P, McCreight C R, et al. Development and testing of a 1024×1024 pixel Si: As IBC detector for SOFIA-like applications[C]//Airborne Telescope Systems Ⅱ of SPIE, 2003, 4857: 155-165.
[29] Hogue H H, Guptill M L, Reynolds D, et al. Space mid-IR detectors from DRS[C]//IR Space Telescopes and Instruments, 2003, 4850: 880-889.
[30] Adams J D, Herter T L, Keller L D, et al. Testing of mid-infrared detector arrays for FORCAST[C]//Optical and Infrared Detectors for Astronomy of SPIE, 2004, 5499: 442-451.
[31] Love P J, Hoffman A W, Lum N A, et al. 1024×1024 Si: As IBC detector arrays for JWST MIRI[C]//Focal Plane Arrays for Space Telescopes Ⅱ of SPIE, 2005, 5902: 58-66.
[32] Mainzer A K, Hong J, Stapelbroek M G, et al. A new large-well 1024×1024 Si: As detector for the mid-infrared[C]//Infrared and Photoelectronic Imagers and Detector Devices of SPIE, 2005, 5881: 253-260.
[33] Mainzer A, Larsen M, Stapelbroek M G, et al. Characterization of flight detector arrays for the wide-field infrared survey explorer[C]//High Energy, Optical, and Infrared Detectors for Astronomy Ⅲ of SPIE, 2008, 7021: 302-313.
[34] Ives D, Finger G, Jakob G, et al. AQUARIUS: the next generation mid-IR detector for ground-based astronomy[C]//High Energy, Optical, and Infrared Detectors for Astronomy Ⅴ of SPIE, 2012, 8453: 296-308.
[35] Reynolds D B, Seib D H, Stetson S B, et al. Blocked impurity band hybrid infrared focal plane arrays for astronomy[J]. IEEE Transactions on Nuclear Science, 1989, 36(1): 857-862. DOI: 10.1109/23.34565
[36] Petroff M D, Stapelbroek M G. Blocked Impurity Band Detectors, Radiation Hard, High Performance LWIR Detectors[C]//Proceedings, IRIS Specialty Group on Infrared Detectors, 1980: 48-62.
[37] Mainzer A, Larsen M, Stapelbroek M G, et al. Characterization of flight detector arrays for the wide-field infrared survey explorer[C]//High Energy, Optical, and Infrared Detectors for Astronomy Ⅲ of SPIE, 2008, 7021: 302-313.
[38] Ando K J, Hoffman A W, Love P J, et al. Development of Si: As impurity band conduction (IBC) detectors for mid-infrared applications[C]//Infrared Technology and Applications XXIX, 2003, 5074: 648-657.
[39] LIAO K, LI N, LIU X, et al. Ion-implanted Si: P blocked-impurity-band photodetectors for far-infrared and terahertz radiation detection[C]//International Symposium on Photoelectronic Detection and Imaging on Terahertz Technologies and Applications of SPIE, 2013, 8909: 257-265.
[40] Hogue H, Atkins E, Reynolds D, et al. Update on blocked impurity band detector technology from DRS[C]//Detectors and Imaging Devices: Infrared, Focal Plane, Single Photon. International Society for Optics and Photonics, 2010, 7780: 778004.
[41] Sclar N. Properties of doped silicon and germanium infrared detectors[J]. Progress in Quantum Electronics, 1984, 9(3): 149-257. DOI: 10.1016/0079-6727(84)90001-6
[42] Kleinhans W A, Petroff M D, Stapelbroek M G. Hybrid Si: As BIBIB Detector Arrays[C/OL]//Proc. of IRIS Specialty Group on Infrared Detectors, 1984: https://www.researchgate.net/profile/George-Gull/publication/234236444_Improved_SiAs_BIBIB_Back-Illuminated_Blocked-Impurity-Band_hybrid_arrays/links/0f317537a14b676810000000/Improved-SiAs-BIBIB-Back-Illuminated-Blocked-Impurity-Band-hybrid-arrays.pdf.
[43] Fowler A M, Joyce R R. Status of the NOAO evaluation of the Hughes 20x64 Si: As impurity band conduction array[C]//Instrumentation in Astronomy VⅡ, 1990, 1235: 151-159.
[44] Larsen M F, Sargent S D, Tansock Jr J J. On-orbit goniometric calibration for the SPIRIT Ⅲ radiometer[C]//Signal and Data Processing of Small Targets of SPIE, 1998, 3373: 32-43.
[45] Hoffman A W, Love P J, Ando K J, et al. Large infrared and visible arrays for low-background applications: an overview of current developments at Raytheon[C]//Optical and Infrared Detectors for Astronomy, 2004, 5499: 240-249.
[46] Mainzer A K, Hogue H, Stapelbroek M, et al. Characterization of a megapixel mid-infrared array for high background applications[C]//High Energy, Optical, and Infrared Detectors for Astronomy Ⅲ, 2008, 7021: 70210T.
[47] Hogue H H, Mattson R B, Stapelbroek M G, et al. Focal plane detectors for the WISE 12-and 23-µm bands[C]//Infrared Systems and Photoelectronic Technology Ⅱ of SPIE, 2007, 6660: 194-202.
[48] Mainzer A K, Hong J, Stapelbroek M G, et al. A new large-well 1024×1024 Si: As detector for the mid-infrared[C]//Infrared and Photoelectronic Imagers and Detector Devices, 2005, 5881: 58810Y.
[49] McMurray Jr R E, Johnson R R, McCreight C R, et al. Si: As IBC array performance for SIRTF/IRAC[C]//Infrared Spaceborne Remote Sensing Ⅷ of SPIE, 2000, 4131: 62-69.
[50] Ando K J, Hoffman A W, Love P J, et al. Development of Si: As impurity band conduction (IBC) detectors for mid-infrared applications[C]//Infrared Technology and Applications XXIX, 2003, 5074: 648-657.
[51] Starr B, Mears L, Fulk C, et al. RVS large format arrays for astronomy[C]//High Energy, Optical, and Infrared Detectors for Astronomy Ⅶ of SPIE, 2016, 9915: 929-942.
[52] Miyata T, Sako S, Nakamura T, et al. Development of a new mid-infrared instrument for the TAO 6.5-m Telescope[C]//Ground-based and Airborne Instrumentation for Astronomy Ⅲ, 2010, 7735: 77353P.
[53] Stacey G J, Hayward T L, Latvakoski H M, et al. KWIC: a widefield mid-infrared array camera/spectrometer for the KAO[C]//Infrared Detectors and Instrumentation, 1993, 1946: 238-248.
[54] Van Cleve J E, Herter T L, Butturini R, et al. Evaluation of Si: As and Si: Sb blocked-impurity-band detectors for SIRTF and WIRE[C]//Infrared Spaceborne Remote Sensing Ⅲ of SPIE, 1995, 2553: 502-513.
[55] Dotson J L, McKelvey M, McMurray Jr R, et al. Cryogenic testing of a 1024×1024 Si: As array for WISE[C]//Focal Plane Arrays for Space Telescopes Ⅲ, 2007, 6690: 66900F.
[56] WANG C, LI N, DAI N, et al. High performance infrared detectors compatible with CMOS-circuit process[J]. Chinese Physics B, 2021, 30(5): 050702. DOI: 10.1088/1674-1056/abd6fb
[57] ZHU H, ZHU J, HU W, et al. Temperature-sensitive mechanism for silicon blocked-impurity-band photodetectors[J]. Applied Physics Letters, 2021, 119(19): 191104. DOI: 10.1063/5.0065468
[58] ZHU H, ZHU J, XU H, et al. Design and fabrication of plasmonic tuned THz detectors by periodic hole structures[J]. Infrared Physics & Technology, 2019, 99: 45-48.
[59] DENG K, ZHANG K, LI Q, et al. High-operating temperature far-infrared Si: Ga blocked-impurity-band detectors[J]. Applied Physics Letters, 2022, 120(21): 211103. DOI: 10.1063/5.0092774
[60] CHEN Y, TONG W, WANG B, et al. The absorption enhancement effect of metal gratings integrated Silicon-based Blocked-Impurity-Band (BIB) terahertz detectors[C]//2021 International Conference on Numerical Simulation of Optoelectronic Devices (NUSOD) of IEEE, 2021: 43-44.
[61] XIAO Y, ZHU H, DENG K, et al. Progress and challenges in blocked impurity band infrared detectors for space-based astronomy[J]. Science China Physics, Mechanics & Astronomy, 2022, 65(8): 1-17.
[62] Herter T L, Hayward T L, Houck J R, et al. Mid-and far-infrared hybrid focal plane arrays for astronomy[C]//Infrared Astronomical Instrumentation of SPIE, 1998, 3354: 109-115.
[63] Stapelbroek M G, Hogue H H, Atkins E W, et al. Silicon for visible-to-VLWIR photon detection[C]//Infrared Technology and Applications XXIX, 2003, 5074: 166-172.
[64] Bamberg J A, Zaun N H. Design and performance of the cryogenic focal plane optics assembly for the Infrared Astronomical Satellite (IRAS)[C]//Cryogenic Optical Systems and Instruments Ⅰ of SPIE, 1985, 509: 94-102.
[65] Werner M W, Roellig T L, Low F J, et al. The Spitzer space telescope mission[J]. The Astrophysical Journal Supplement Series, 2004, 154(1): 1. DOI: 10.1086/422992
[66] Mainzer A K, Eisenhardt P, Wright E L, et al. Preliminary design of the wide-field infrared survey explorer (WISE)[C]//UV/Optical/IR Space Telescopes: Innovative Technologies and Concepts Ⅱ Of SPIE, 2005, 5899: 262-273.
[67] Gardner, J.P., Mather, J.C., Clampin, M., et al. The James Webb Space telescope[J]. Space Science Reviews, 2006, 123(4): 485-606. DOI: 10.1007/s11214-006-8315-7
[68] Huffman J E. Infrared detectors for 2-to 220-um astronomy[C]//Infrared Detectors: State of the Art Ⅱ, 1994, 2274: 157-169.
[69] Herter T, Stacey G, Gull G, et al. FORCAST: a WIDE-field infrared camera for SOFIA[C]//American Astronomical Society Meeting Abstracts, 1997, 191: 09.02.
[70] Mather J C, Cheng E S, Eplee Jr R E, et al. A preliminary measurement of the cosmic microwave background spectrum by the Cosmic Background Explorer (COBE) satellite[J]. The Astrophysical Journal, 1990, 354: L37-L40. DOI: 10.1086/185717
[71] Rauter P, Fromherz T, Winnerl S, et al. Terahertz Si: B blocked-impurity-band detectors defined by nonepitaxial methods[J]. Applied Physics Letters, 2008, 93(26): 261104. DOI: 10.1063/1.3059559
[72] Mary W. Jackson NASA Headquarters. NASA Missions[DB/OL]. https://www.nasa.gov/missions.