Spectrum Aggregation for 6G: Lessons from 5G Carrier Aggregation and Dual Connectivity

1

Spectrum Aggregation for 6G: Lessons from 5G Carrier Aggregation and Dual Connectivity Xingqin Lin

Abstract—Spectrum aggregation has been a key enabler of LTE and 5G capacity growth, but it will become even more fundamental in 6G as networks expand across low bands, existing mid bands, new upper-mid/centimetric bands, and millimeter wave bands. This article examines how 5G carrier aggregation (CA) and dual connectivity (DC) inform the design of 6G spectrum aggregation. We argue that, while DC was instrumental in accelerating early non-standalone 5G deployment, it also introduced architectural fragmentation and long-term migration complexity. In contrast, CA provides a cleaner and more scalable foundation for multi-band operation in standalone 6G. Building on lessons from 5G, we advocate enhanced CA as the preferred 6G aggregation framework and point out the corresponding key enhancement directions.

I. INTRODUCTION Spectrum aggregation has been one of the most important techniques in modern cellular systems. Starting from long-term evolution (LTE) carrier aggregation (CA) and continuing into 5G new radio (NR) developed by the 3rd generation partnership project (3GPP), the ability to combine multiple spectrum resources has enabled operators to increase peak rates, improve cell-edge throughput, and make practical use of fragmented spectrum holdings [1]. In 5G, spectrum aggregation was largely introduced as a capacity enhancement and deploymentenabling feature [2]. In 6G, spectrum aggregation is likely to become an even more important capability, because 6G will need to operate across a multi-layer spectrum fabric spanning legacy low bands, existing mid bands, new uppermid/centimetric bands, millimeter-wave (mmWave) bands, and maybe eventually sub-THz spectrum (though not part of day-1

6G) [3][4]. This makes spectrum aggregation a foundational 6G design requirement [5]. The demand driver is clear: 6G is expected to support not only continued mobile broadband growth, but also immersive extended reality (XR), artificial intelligence (AI) generated traffic, massive digital twins, integrated sensing, and robotics. In 6G, sub-1 GHz bands will remain essential for nationwide coverage, deep indoor reach, rural connectivity, and low-power internet of things (IoT). Existing mid bands below 6 GHz will continue to provide the main wide-area capacity layer, benefiting from mature massive multiple-input multiple-output (MIMO) deployments and broad device support. Upper-mid bands, such as the upper 6 GHz and 7-15 GHz range, are emerging as the most important new 6G candidates because they offer more bandwidth than today’s mid bands while still having propagation characteristics that may support macrocellular deployment with advanced antenna arrays [6]. Therefore, world radiocommunication conference 2027 (WRC27) studies cover bands such as 4.4-4.8 GHz, 7.125-8.4 GHz, and 14.8-15.35 GHz, while many countries are also considering the upper 6 GHz band for international mobile telecommunications (IMT) usage [7]. At higher frequencies, mmWave will provide localized extreme capacity rather than wide-area coverage. The potential sub-THz bands, such as Wband and D-band, may provide extremely large bandwidths and potentially Tbps-class links, but with limited coverage and mobility. These higher frequency bands are therefore more suitable for niche scenarios such as extreme local connectivity, device-to-device links, wireless data-center connectivity, and fronthaul/backhaul. The implication is that no single spectrum

Spectrum layer

Role in 6G

Strength

Limitation

Sub-1 GHz

Coverage, rural, deep indoor, low-power IoT

Excellent propagation

Limited bandwidth

1-6 GHz mid band

Wide-area macro capacity

Upper 6 GHz / 7-15 GHz

New wide-area 6G capacity layer

Strong ecosystem, massive MIMO maturity Wider channels with more macro relevance than mmWave

24-52.6 GHz Dense urban, venues, Very large bandwidth mmWave fixed wireless access Above 52.6 GHz Extreme local Very wide bandwidth, mmWave and subcapacity, Xhaul, fixed extreme rates THz* links *Note: This spectrum layer is not expected to be part of day-1 6G.

Insufficient alone for 6G capacity

Aggregation implication Anchor mobility and reliability Aggregate with uppermid bands for capacity scaling

More challenging propagation and coexistence Blockage and limited coverage

Opportunistic capacity boost

Limited mobility and maturity

Specialized highcapacity layer

Candidate main 6G expansion layer

Table 1: Comparative summary of 6G spectrum layers and their aggregation implications.

2 layer can satisfy all 6G requirements [8]. Therefore, 6G must be designed from the beginning to aggregate spectrum across layers in different frequency ranges (FRs), as summarized in Table 1. 5G already provides two major tools for this purpose: CA and DC [9]. CA combines multiple component carriers under a common serving-cell framework, with one primary cell (PCell) and one or more secondary cells (SCells) [10]. DC allows a user equipment (UE) to connect to two cell groups, usually controlled by different network nodes, such as an LTE master node and an NR secondary node in early non-standalone 5G deployments [11]. These mechanisms have been deployed with varying degrees of success. This article draws lessons from 5G CA and DC and discusses how these lessons can guide 6G spectrum aggregation design. The central thesis is that 6G should preserve the commercial and implementation benefits of CA, while addressing the shortcomings of 5G CA. The article first reviews the basics of 5G CA and DC, and summarizes the lessons learnt. Then, this article argues that 6G should adopt enhanced CA as the primary spectrum aggregation framework in 6G and points out key enhancement directions for 6G CA. II. FUNDAMENTALS OF 5G CA AND DC CA and DC are the two most important mechanisms for aggregating radio resources across carriers, frequency bands, and network nodes in 5G. Both allow a UE to use more than one serving carrier, but they are based on different architectural assumptions. CA is primarily a multi-carrier aggregation mechanism, while DC is primarily a multi-cell-group aggregation mechanism. Understanding their differences is essential for designing 6G spectrum aggregation, because many 6G proposals can be viewed as modifications or extensions of these two 5G mechanisms. A. CA in 5G CA is a fundamental 3GPP mechanism for increasing user throughput, improving spectrum utilization, and enabling flexible operation across fragmented spectrum holdings. Instead of serving a UE on a single carrier, CA allows the network to configure multiple component carriers for the same UE and schedule data over them in a coordinated manner. Conceptually, a component carrier is an individually configured carrier with its own bandwidth, numerology, duplexing mode, bandwidth parts, reference signals, control-resource configuration, and scheduling resources. As illustrated in Figure 1, CA combines multiple such carriers under one radio resource control (RRC) connection and one medium access control (MAC) entity, so that the UE experiences the aggregated bandwidth as a larger logical pipe while the 5G node B (gNB) retains per-carrier scheduling flexibility. This design is particularly useful in 5G NR because deployments may combine low-band coverage carriers, mid-band capacity carriers, and high-band mmWave carriers, each with different propagation, bandwidth, and deployment characteristics. A central feature of CA is the distinction between PCell and SCell. The PCell provides the anchor for the UE’s RRC connection and essential control-plane operation. In 5G NR, the PCell belongs to a primary cell group and supports functions such as initial access, radio link monitoring, and key control

procedures. Additional carriers are configured as SCells, which mainly provide extra user-plane capacity. The network may add, release, activate, or deactivate SCells depending on traffic demand, UE capability, channel conditions, and energyefficiency considerations. This PCell/SCell structure allows NR CA to separate connection robustness from opportunistic capacity expansion. From an RRC perspective, CA is configured through dedicated signaling. The gNB informs the UE which serving cells are configured, what carrier frequencies and bandwidths are used, which bandwidth parts are available, how downlink and uplink control are arranged, and which measurement and reporting procedures apply. The configuration may include multiple SCells, but not all configured SCells need to be active at a given time. This separation between configuration and activation is important: configuration prepares the UE with the required parameters, while activation enables rapid use of the SCell without repeating the full RRC reconfiguration procedure. The MAC layer then controls SCell activation and deactivation using MAC control elements, enabling faster adaptation than RRC-only control. At the MAC layer, NR CA is designed around a single MAC entity that can multiplex logical-channel data across transport blocks transmitted on different serving cells. This means that packet data from higher layers is not statically tied to one carrier. Instead, the scheduler can dynamically decide whether to transmit on the PCell, one or more SCells, or a combination of carriers. This allows the gNB to exploit instantaneous channel quality, load, interference, and traffic priorities across carriers. For example, the scheduler may use a low-band PCell to maintain robust coverage and control, while using a wide mid-band or mmWave SCell for bursty high-throughput downlink traffic [12]. In the downlink, each activated serving cell may have its own physical downlink shared channel (PDSCH) transmissions, hybrid automatic repeat request (HARQ) processes, channel state information (CSI) reporting configuration, and scheduling behavior. Scheduling may be performed using self-carrier scheduling, where the physical downlink control channel (PDCCH) on a carrier schedules data on the same carrier, or cross-carrier scheduling, where control information on one serving cell schedules data on another serving cell. Crosscarrier scheduling is useful when one carrier has more reliable or more available control resources than another, or when the network wants to reduce control-channel monitoring burden on selected carriers. The downlink control information includes carrier-indication information when cross-carrier scheduling is configured, allowing the UE to determine which scheduled cell the grant applies to. In the uplink, CA introduces additional design constraints because UE transmit power, radio frequency (RF) chain capability, simultaneous transmission capability, and intermodulation constraints must be respected. A UE may support different combinations of uplink carriers depending on its capability and the standardized band combination. In some deployments, downlink CA is configured more aggressively than uplink CA because the downlink capacity demand is larger and UE uplink transmission across multiple bands is more challenging. 5G NR also supports supplementary uplink concepts, which may be combined with broader multi-carrier operation to improve uplink coverage.

3

Figure 1: An illustration of radio protocol architecture for 5G downlink CA [9].

HARQ operation is also carrier specific. Each serving cell can maintain its own HARQ timing and processes, while HARQ feedback may be multiplexed and transmitted over configured uplink control resources. This creates a key control design challenge: as the number of aggregated carriers increases, the volume of downlink assignments and HARQACK feedback can increase substantially. 3GPP therefore specifies procedures for HARQ-ACK codebook construction, physical uplink control channel (PUCCH) resource selection, and uplink control information multiplexing to support multicell operation without overwhelming uplink control capacity. CSI is another important aspect of CA. Since each carrier may experience different path loss, fading, interference, and traffic load, CSI acquisition is generally carrier specific. The UE may be configured with CSI reference signal (CSI-RS) resources and reporting configurations on different serving cells, enabling the scheduler to select modulation and coding schemes, MIMO precoders, and transmission layers independently per carrier. NR CA supports several deployment forms. In intra-band contiguous CA, the aggregated component carriers are adjacent in frequency and may be handled efficiently by the RF front end. In intra-band non-contiguous CA, carriers are in the same band but separated in frequency, allowing operators to combine fragmented holdings. In inter-band CA, carriers from different bands are aggregated, for example combining a low-band coverage carrier with a mid-band capacity carrier. Inter-band CA provides powerful coverage-capacity tradeoffs but is more demanding for UE RF design, power consumption, and coexistence. 3GPP therefore defines CA band combinations and UE capability signaling so that the network only configures combinations supported by the UE and deployment. CA also introduces important design tradeoffs. First, UE power consumption increases when the device monitors multiple carriers, maintains multiple transmitter or receiver chains, or performs CSI measurements across several cells. SCell dormancy and activation/deactivation are therefore important tools for balancing throughput and energy efficiency. Second, control overhead can grow with the number of carriers,

especially for PDCCH monitoring, CSI reporting, and HARQ feedback. Third, mobility and coverage management become more complex because the best carrier for capacity may not be the best carrier for reliability. Finally, inter-band CA places significant demands on RF front-end design, including filters, power amplifiers, antenna tuning, and simultaneous transmit/receive constraints. B. DC in 5G DC is a 3GPP multi-connectivity framework that allows a UE to simultaneously use radio resources from two different radio access nodes. While CA combines multiple component carriers under relatively tight coordination within one scheduling framework, DC enables a looser but more flexible form of aggregation across two nodes, two cell groups, and, in many deployments, two radio access technologies (RATs). In 5G, this concept is generalized as multi-radio dual connectivity (MRDC). The basic architectural idea is shown by the distinction between a master node and a secondary node. The master node provides the UE’s control-plane connection to the core network, while the secondary node provides additional radio resources without having its own independent control-plane connection to the core network for that UE. This organization allows the UE to maintain a robust control anchor through the master node while exploiting additional bandwidth from the secondary node. A particularly important form of DC in early 5G deployment was EN-DC, where an LTE evolved Node B (eNB) acts as the master node and an NR gNB acts as the secondary node. This enabled non-standalone 5G deployments, where LTE provided the mobility and control-plane anchor while NR delivered high-throughput user-plane capacity. 3GPP also defines other MR-DC variants, such as NE-DC and NR-DC, depending on whether the master and secondary nodes are LTE eNBs or NR gNBs. From an architectural perspective, the common principle is the same: one node anchors the control relationship, and the second node contributes additional radio resources.

4

Figure 2: An illustration of radio protocol architecture for ENDC [13].

In MR-DC terminology, the cells associated with the master node form the master cell group (MCG), and the cells associated with the secondary node form the secondary cell group (SCG). The MCG contains the PCell and may additionally contain one or more SCells. The SCG contains the primary secondary cell (PSCell) and may additionally contain one or more SCells. As illustrated in Figure 2, each cell group has its own lower-layer protocol stack, including radio link control (RLC), MAC, and physical-layer resources. This is a key distinction from CA. In CA, multiple serving cells are controlled by one MAC entity and one scheduler within a gNB. In DC, the master node and secondary node have separate MAC entities. Coordination is mainly performed through higher-layer configuration, bearer mapping, and inter-node interfaces. This makes DC more suited for deployments where the aggregated resources are not co-located, do not share the same scheduler, or belong to different RATs. The most important user-plane concept in DC is the radio bearer architecture. 3GPP defines several bearer types depending on where the protocol entities are terminated and how traffic is routed. An MCG bearer uses radio resources only from the MCG. An SCG bearer uses radio resources only from the SCG. A split bearer allows traffic for the same radio bearer to be transmitted through both MCG and SCG resources. In a split bearer, the packet data convergence protocol (PDCP) entity is associated with multiple RLC entities so that packets can be routed over different lower-layer paths. Note that, for split bearers or bearers configured with PDCP duplication, one PDCP entity may be associated with multiple RLC entities. The split-bearer concept is central to the performance benefit of DC. Since PDCP sits above RLC, MAC, and PHY, it can distribute user-plane packets across the master node and secondary node paths. This allows the network to exploit two radio links simultaneously even when the two nodes are independently scheduled. For high-throughput traffic, a split bearer can increase aggregate data rate by using resources from both cell groups. For reliability-sensitive traffic, PDCP duplication can transmit duplicate PDCP protocol data units over two paths, improving robustness against link blockage, interference, or mobility-related degradation. The tradeoff is higher radioresource consumption, since duplicated packets consume capacity on both legs. From a control-plane perspective, the master node remains the main anchor. RRC signaling is primarily controlled by the master node, and the secondary node is added, modified, or

released through coordinated RRC procedures. In DC operation, the UE may receive RRC configurations that establish the SCG, configure the PSCell, assign radio bearers to MCG, SCG, or split operation, and define measurement and reporting behavior for maintaining both links. Mobility management in DC is also more complex than in singleconnectivity operation. Since the UE maintains two cell groups, the network may need to perform MCG mobility, SCG addition, SCG release, or PSCell change. The master node typically coordinates these procedures, using UE measurements and inter-node signaling to determine whether the secondary node should be added, changed, or removed. For example, if a UE moves out of the coverage of an NR secondary node, the network may release the SCG while maintaining the MCG connection. Conversely, if the UE enters good NR coverage, the network may add or reactivate an NR secondary node to increase throughput. Compared with CA, DC relaxes synchronization and scheduler-coordination requirements but introduces additional protocol and network complexity. CA is generally more efficient when carriers are controlled by the same gNB and can be scheduled jointly at the MAC/PHY level. DC, in contrast, is more flexible when resources are distributed across different nodes. However, because the two cell groups have separate MAC entities and potentially different latency, backhaul, and scheduler behavior, split-bearer operation must handle packet reordering, flow control, and path imbalance. PDCP reordering and duplication control therefore become important for maintaining in-sequence delivery and avoiding excessive delay. Another key design issue is inter-node transport. In DC, the master node and secondary node exchange control and userplane information over standardized inter-node interfaces. The quality of this interface affects the achievable performance of split bearers and secondary-node operation. If the inter-node interface has high latency or limited capacity, aggressive packet splitting may cause reordering delay or inefficient resource use. As a result, practical DC performance depends not only on radio conditions but also on transport-network design, scheduler implementation, and bearer-selection policy. In addition, for uplink operation, DC must consider UE power limits and simultaneous transmission capability. A UE transmitting to both MCG and SCG may need to share limited power across two radio links, possibly in different frequency bands. The network therefore needs to configure uplink resources carefully, considering power control, timing advance, RF capability, and maximum permissible exposure constraints. In many deployments, the downlink benefit of DC is more straightforward, while uplink DC requires more careful management of UE capability and power constraints. III. LESSONS FROM 5G CA AND DC 5G CA and DC are both 3GPP mechanisms for increasing the effective radio resources available to a UE, but they solve different deployment problems. In simplified terms, CA is a tightly coordinated carrier-level aggregation mechanism, whereas DC is a more loosely coordinated node-level or cellgroup-level aggregation mechanism. This distinction has important consequences for network architecture, scheduler design, device complexity, mobility robustness, and

5 Aspect

5G Carrier Aggregation Aggregates multiple carriers for one UE

5G Dual Connectivity Aggregates resources from two nodes or cell groups

Component-carrier level

Cell-group / node level

Combining low-band, midband, or TDD/FDD carriers within NR PCell anchors key control procedures

Inter-node dependency

One MAC entity controls multiple serving cells Generally centralized within one gNB Mainly MAC-level multiplexing across carriers No inter-node interface needed for aggregation

EN-DC with LTE anchor and NR secondary node; NR-DC across two NR nodes Master node anchors the control plane SCells in PCG, PSCell and SCells in SCG Separate MAC entities for MCG and SCG Independently performed by master node and secondary node PDCP-level split bearer or duplication Requires coordination between master node and secondary node

Synchronization and coordination

Tighter timing and scheduling coordination

Looser coordination between nodes

Basic purpose Aggregation level Typical 5G use case Control anchor Secondary resource Protocol structure Scheduling User-plane split

RF/device complexity Mobility behavior Uplink considerations Control overhead Energy impact Reliability potential Best deployment fit

SCells

Implication CA is best for spectrum pooling; DC is best for heterogeneous node aggregation CA is more radio-scheduler centric; DC is more architecture and transport centric DC was crucial for early non-standalone 5G, while CA becomes increasingly important in mature standalone 5G CA depends on a robust primary carrier; DC depends on a robust master node CA adds carriers; DC adds another leg CA enables tighter scheduling; DC enables looser but more flexible deployment CA can be more spectrally efficient; DC can exploit non-co-located resources DC can support path diversity and duplication, but with reordering and transport challenges DC performance strongly depends on Xn transport quality and vendor implementation CA is better suited to co-sited or tightly integrated carriers; DC is better for distributed deployments

Requires support for specific Requires support for MR-DC Both are UE-capability limited, but DC can CA band combinations and combinations and possibly impose heavier multi-node mobility and simultaneous multiple radio chains power-control complexity reception/transmission Mobility mainly centered Requires MCG mobility, SCG DC offers flexibility but increases mobility around the PCell and addition/release, PSCell change, management complexity configured SCells and secondary node change Uplink CA limited by UE Uplink DC limited by UE power Downlink gains are easier to commercialize power, RF capability, and sharing across two nodes and than uplink gains for both CA and DC supported band combinations timing alignment More PDCCH monitoring, Additional RRC, PDCP, interCA overhead is mostly radio-control CSI reporting, and HARQ node, and SCG management overhead; DC overhead is both radio and feedback across carriers overhead network-protocol overhead Multiple active carriers SCell activation/deactivation, dormancy, and Maintaining two cell groups can increase UE power SCG management are essential for battery increase UE power significantly consumption efficiency Frequency diversity across Path diversity across nodes; DC has stronger reliability potential when the carriers PDCP duplication possible two legs are sufficiently independent Same-site or tightly CA is commercially attractive where spectrum coordinated multi-band NR EN-DC, or inter-site aggregation is fragmented; DC is attractive where deployment topology is heterogeneous Table 2: 5G CA versus DC: technical differences and implications.

commercial deployment strategy. Table 2 summarizes the main differences between 5G CA and DC and the corresponding implications, which are elaborated below. Several lessons can be learnt from 5G CA and DC experiences. The first lesson is that CA and DC entered commercial 5G with different roles. In early non-standalone 5G deployments, EN-DC was the practical foundation for rapid 5G rollout. LTE provided wide-area coverage, mature mobility, and control-plane anchoring, while NR provided highthroughput capacity on newly deployed mid-band or mmWave carriers. This allowed operators to introduce 5G capacity without immediately deploying a fully standalone 5G core (5GC) and complete NR coverage layer. From a commercial perspective, DC was not only a throughput feature but also an architectural bridge from LTE to NR. By contrast, CA becomes increasingly important as NR deployments mature. Once operators deploy standalone NR and multiple NR carriers, the

value of CA grows because it allows fragmented spectrum assets to be pooled efficiently. For example, an operator may combine a low-band NR carrier for coverage with a mid-band TDD carrier for capacity, or aggregate multiple mid-band carriers to increase peak and cell-edge throughput. In mature 5G networks, CA is often the cleaner mechanism for bandwidth scaling when the carriers are controlled by the same gNB. A second lesson is that CA is generally more schedulerefficient, while DC is more deployment-flexible. Since CA is typically controlled within one gNB scheduling framework, the scheduler can jointly consider channel quality, buffer status, HARQ state, traffic priority, and carrier load across multiple serving cells. This makes CA attractive for maximizing spectral efficiency and reducing coordination delay. DC, however, allows the network to combine resources that may not be controlled by the same scheduler, such as LTE and NR nodes, macro and small-cell nodes, or geographically separated NR

6 nodes. This makes DC more flexible, but the flexibility comes at the cost of more complex bearer management, inter-node coordination, and packet reordering. A third lesson is that control overhead and UE power consumption become key design constraints, while downlink aggregation is easier to exploit than uplink aggregation. Aggregating more carriers or maintaining two cell groups increases PDCCH monitoring, CSI measurement and reporting, HARQ feedback, RF-chain activity, and baseband processing. Commercially, the objective is not to keep every carrier or cell group active all the time, but to activate additional resources only when the expected traffic and radio conditions justify the cost. In the downlink, the network transmits from infrastructure nodes with fewer power and RF constraints. In the uplink, the UE must share limited transmit power across carriers or nodes, and simultaneous transmission may be limited by RF architecture, thermal constraints, exposure regulations, and intermodulation issues. Therefore, many commercial deployments initially emphasize downlink CA or downlink DC gains. Uplink CA and uplink DC are valuable for highthroughput uplink services, but they are more constrained and require careful power-control and scheduling design. A fourth lesson is that DC is highly sensitive to transport quality. In split-bearer operation, packets may traverse both the master node and the secondary node paths. If the two paths have different latency, congestion, or scheduling behavior, PDCP reordering delay may reduce the expected gain. Similarly, if the inter-node interface is slow or congested, the secondary node may not be used efficiently. This means that DC performance is not only determined by radio bandwidth but also depends on backhaul latency, inter-node flow control, vendor interoperability, scheduler coordination, and radio resource management (RRM) policies. Last but not the least, while EN-DC was commercially effective for fast 5G launch, it also made DC carry two roles at once: a useful performance feature and a temporary architectural bridge. As a result, operators had to manage LTE/NR interworking, evolved packet core (EPC) to 5GC migration, non-standalone and standalone device behaviors, separate optimization paths for EN-DC and NR-only operation, and delayed use of native 5GC capabilities [14]. By contrast, CA is a cleaner long-term spectrum-scaling tool because it aggregates carriers inside a native NR system under tighter scheduler control, rather than relying on another legacy node as the control anchor. The 6G lesson is therefore that DC should not be the migration shortcut. 6G should start from a standalone architecture with 6G radio access available across the needed spectrum bands, so that spectrum aggregation can be designed natively through CA-like multi-band operation. IV. 6G SPECTRUM AGGREGATION OUTLOOK The key design question for 6G spectrum aggregation is not only how to combine more carriers, but how to do so without repeating the architectural fragmentation experienced in 5G. This section argues that enhanced CA should serve as the preferred baseline for 6G spectrum aggregation, followed by key enhancement directions for 6G CA.

A. Enhanced CA as the Preferred Baseline A central question for 6G spectrum aggregation is whether the system should continue to rely on two largely separate mechanisms, namely CA and DC, or whether 6G should converge toward a simpler and more unified aggregation framework. The 5G experience provides a clear lesson: too many architectural options may accelerate early deployment in the short term, but they also create long-term fragmentation, device complexity, network optimization burden, and migration difficulty. 6G should narrow migration and architecture options early to reduce complexity across UEs, radio access network (RAN), and core network. From this perspective, enhanced CA should be the preferred baseline method for 6G spectrum aggregation. This does not mean that every deployment scenario must be forced into a tightly synchronized single-scheduler model. Rather, it means that 6G should first determine how far an enhanced CA framework can be extended before introducing a separate 6G6G DC architecture. Recent 3GPP discussions already point in this direction: For non-collocated FR1 and FR2 aggregation with backhaul latency up to 10 ms, both 6G-6G DC with two cell groups and 6G-6G CA with two PUCCH groups and separate timing advance per group have been identified as capable of supporting imperfect-backhaul aggregation [15]. Importantly, the physical-layer functionalities of the two solutions are similar. This weakens the case for introducing DC as a parallel baseline mechanism for 6G-6G aggregation: if the same lower-layer functions can be provided by CA with two PUCCH groups and separate timing advance, then CA offers a cleaner path with less architectural duplication. The distinction becomes clearer when considering three candidate aggregation modes. 6G-6G CA with a single PUCCH group represents the tightest form of aggregation. It is attractive when the FR1 and FR2 resources are sufficiently coordinated, because uplink control for FR2 downlink transmissions can be sent on FR1. This may improve uplink control coverage, especially when FR2 uplink coverage is weak. However, for non-collocated deployments with imperfect backhaul, CA with a single PUCCH group also creates a timing challenge: HARQACK for FR2 may need to traverse backhaul and follow a longer rescheduling timeline, increasing HARQ round-trip time (RTT). If the HARQ RTT becomes too large, the scheduler may run out of available HARQ processes, causing HARQ process starvation and downlink throughput degradation. 6G-6G CA with two PUCCH groups provides a more flexible middle ground. Each group may maintain its own uplink control resources and timing advance, allowing FR1 and FR2 legs to operate with more relaxed coordination. This is especially relevant for non-collocated deployments where the two transmission points do not share ideal backhaul or tight timing. Compared with single-PUCCH-group CA, two-PUCCH-group CA reduces tight cross-carrier timing dependencies and can avoid sending every control response through the more favorable but potentially backhaul-constrained FR1 path. Compared with DC, it can provide similar physical-layer functionality while preserving the CA abstraction and avoiding a separate cell-group architecture.

7

Figure 3: An illustration of key enhancement directions for 6G CA.

6G-6G DC with two cell groups provides the loosest coordination model. It may still be useful when the two nodes are sufficiently independent that CA assumptions become too restrictive. However, DC also brings the burdens of separate cell groups, separate lower-layer protocol entities, inter-node coordination, bearer management, and more complex mobility and failure handling. The non-standalone 5G experience further warns against using DC as a default migration shortcut. In particular, many operators found it difficult to migrate from non-standalone to standalone 5G to realize 5GC-based features [14]. This is the type of architectural split that 6G should avoid. DC-like operation should be reserved only for cases where independent cell-group operation provides a clearly justified benefit that cannot be achieved within enhanced CA. B. Key Enhancement Directions for 6G CA To make CA suitable as the primary 6G spectrumaggregation framework, 6G CA should address the main limitations observed in 5G CA and extend the framework. As illustrated in Figure 3, five enhancement areas are particularly important. 1) CA for collocated and non-collocated deployments 6G CA framework should support multiple coordination levels. In 5G, CA is mainly associated with tightly coordinated carriers under a common scheduling framework, while DC is used when resources belong to different nodes or cell groups. In 6G, this separation should be relaxed. For non-collocated FR1-FR2 aggregation with imperfect backhaul, both 6G-6G DC with two cell groups and 6G-6G CA with two PUCCH groups and separate timing advance per group can support the target deployment scenario, with similar physical-layer functionalities. This suggests that enhanced CA can cover scenarios that previously would have been mapped to DC.

The 6G CA framework should therefore support a continuum of coordination modes. For tightly coordinated or collocated carriers, a single PUCCH group may be efficient and may even improve FR2 downlink feedback coverage by transmitting HARQ-ACK on a more robust FR1 uplink. For non-collocated carriers with backhaul latency up to 10 ms, however, a single PUCCH group may increase HARQ round-trip time and lead to HARQ process starvation. In such cases, two PUCCH groups with separate timing advance provide a more robust CA mode. This allows 6G to use a common CA framework to adapt the degree of coordination to the deployment. 2) Fast and energy-efficient carrier activation Reducing the latency and energy cost of SCell activation, deactivation, and addition is another important enhancement direction. One of the key shortcomings of 5G CA is that SCell activation can be too slow for bursty traffic. If a secondary carrier is activated only after the UE buffer has already drained, the additional carrier provides little user-perceived benefit. This issue will become more important in 6G, where XR, AI traffic, sensing updates, and short high-rate transfers may produce highly bursty traffic patterns. 6G CA should therefore support both fast activation and near-instant activation. Fast activation applies when the UE must acquire synchronization and measurements on the additional carrier. Near-instant activation applies when the UE can derive timing, synchronization, or other properties of one carrier from an already active carrier based on network assistance, quasi-co-location relationships, common timing assumptions, or preconfigured carrier groups. This should be coupled with energy-efficient dormant states, allowing the UE to avoid continuous monitoring of secondary carriers while still being able to activate them quickly when traffic demand arises. The objective is to make CA responsive enough for short data bursts without forcing all carriers to remain active all the time.

8 3) Flexible downlink/uplink association and unified uplink switching A third enhancement direction is to enable downlink/uplink decoupling in 6G CA. In many 6G deployments, the best downlink carrier will not be the best uplink carrier. For example, upper-mid-band or FR2 carriers may provide large downlink bandwidth, while low-band or lower-mid-band carriers may provide better uplink coverage, more reliable uplink control, and lower UE power burden. A rigid assumption that downlink and uplink should remain paired on the same carrier would limit the value of multi-band aggregation. 6G CA should therefore support flexible association among downlink data, uplink data, and uplink control across different carriers. This includes transmitting downlink data on a highcapacity carrier while sending HARQ-ACK, scheduling requests, CSI, or even uplink data on a more robust lowerfrequency carrier. It also requires a unified treatment of uplink carrier switching, sounding reference signal (SRS) switching, and uplink transmit-chain switching. Rather than defining separate mechanisms for each case, 6G should integrate these functions into the CA framework, with clear rules for power control, timing advance, pathloss reference, and UE capability signaling. 4) Non-PCell-centric mobility and failure recovery Another important enhancement direction is to relax the rigid PCell-centric design in 5G CA. In 5G, changing the PCell triggers handover, even when the desired operation is essentially a role change between an already configured PCell and SCell. This causes unnecessary interruption and signaling overhead. Similarly, PCell radio link failure triggers RRC reestablishment and causes all SCells to be dropped, even when some secondary carriers remain usable. These behaviors are inefficient for a 6G system that may rely on multiple carriers for coverage, capacity, control, and resilience. 6G CA should therefore support more flexible PCell and SCell role management. A PCell role should be changeable within a configured multi-carrier context without always requiring a full handover. Failure handling should also become carrier-group-aware rather than strictly PCell-based. If one carrier fails but other configured carriers remain reliable, the UE should be able to recover through a surviving carrier, preserve part of the aggregation configuration, and avoid unnecessary teardown of the full multi-carrier context. This would make CA more resilient for upper-mid-band, FR1-FR2, and non-collocated deployments, where different carriers may experience different blockage, interference, or mobility conditions. 5) Carrier grouping and virtual-cell abstraction Introducing a carrier-grouping or virtual-cell abstraction for tightly coupled fragmented spectrum is another enhancement direction. In conventional CA, each carrier is modeled largely as an independent serving cell, which can duplicate SSB, RRM, PDCCH, and mobility overhead. This is inefficient when multiple fragmented carriers are collocated, close in frequency, and share common properties such as timing, numerology, duplexing pattern, frame structure, and MIMO assumptions. For such deployments, 6G may expose multiple physical carriers as one logical carrier group or virtual cell. This abstraction could support one synchronization/reference carrier, reduced or SSB-less operation on secondary physical

carriers, common configuration across the group, one downlink control information (DCI) scheduling resources across multiple physical carriers, common RRM treatment, and simplified mobility. However, this concept should complement rather than replace CA. Enhanced CA should remain the general framework for most spectrum aggregation scenarios, while the virtual-cell abstraction can be used when the network can guarantee sufficiently common properties across the grouped carriers. V. CONCLUSIONS Spectrum aggregation will be a foundational capability for 6G. The 5G experience with CA and DC offers an important lesson: while DC was useful for accelerating non-standalone 5G deployment, it also introduced architectural fragmentation and long-term migration complexity. This article has argued that enhanced CA should be the preferred baseline for 6G spectrum aggregation. Ultimately, the goal of 6G spectrum aggregation should be to transform fragmented and diverse spectrum assets into an efficient and scalable radio access system. A CA-centric design can help 6G achieve this goal while supporting architectural simplification and long-term deployment flexibility. REFERENCES [1]

Ericsson, “6G spectrum - enabling the future mobile life beyond 2030,” white paper, May 2024. [2] E. Dahlman, S. Parkvall, and J. Skold, “5G NR: The next generation wireless access technology,” Academic Press, 2020. [3] W. Chen et al., “5G-Advanced toward 6G: Past, present, and future,” IEEE Journal on Selected Areas in Communications, vol. 41, no. 6, pp. 1592-1619, Jun. 2023. [4] ITU-R, “Framework and overall objectives of the future development of IMT for 2030 and beyond,” Recommendation ITU-R M.2160, Nov. 2023. [5] 3GPP TR 38.914, “Study on 6G scenarios and requirements,” Version 0.4.0, Mar. 2026. [6] A. Bazzi et al., “Upper mid-band spectrum for 6G: vision, opportunity and challenges,” IEEE Communications Magazine, vol. 64, no. 1, pp. 206212, Jan. 2026. [7] ITU-R, “Resolution 256 (WRC-23): Sharing and compatibility studies and development of technical conditions for the use of IMT in the frequency bands 4400–4800 MHz, 7125–8400 MHz, and 14.8–15.35 GHz,” WRC-23, 2023. [8] X. Lin, “A tale of two mobile generations: 5G-Advanced and 6G in 3GPP release 20,” IEEE Communications Standards Magazine, early access, Oct. 2025. [9] 3GPP TS 38.300, “NR and NG-RAN overall description; Stage-2,” V19.2.0, Mar. 2026. [10] W. Ye et al., “Dissecting carrier aggregation in 5G networks: Measurement, QoE implications and prediction.” Proceedings of the ACM SIGCOMM 2024 Conference, pp. 340-357, 2024. [11] M. Agiwal et al., “A survey on 4G–5G dual connectivity: Road to 5G implementation,” IEEE Access, vol. 9, pp. 16193-16210, 2021. [12] X. Lin, “A 3GPP perspective on spectrum sharing for 5G-to-6G migration: From LTE-NR DSS lessons to MRSS design,” arXiv:2604.26853, Apr. 2026. [13] 3GPP TS 37.340, “E-UTRA and NR; multi-connectivity; overall description; Stage 2,” V19.2.0, Mar. 2026. [14] NGMN Alliance, “6G architecture and migration options - an operator view,” Version 1.0, Jun. 2026. [15] 3GPP RAN WG1, “LS on 6G-6G spectrum aggregation for noncollocated deployment scenarios,” RP-260848, Jun. 2026.